# Operation reference Browse request fields, response fields, and required scopes for every REST, CLI, and MCP operation. This reference is generated from the server's registered inputs and typed service results. Call each operation with `POST /api/agent/OPERATION_NAME`. See [API setup](/docs/reference/api) for authentication, examples, pagination, and error handling. The [OpenAPI specification](/openapi.json) includes nested schemas and constraints; the tables below summarize top-level fields. An older deployment may expose fewer operations. Use live discovery on the target server. All successful REST responses wrap the documented result in `{ "data": ... }`. Required fields describe the serialized response; null is distinct from an omitted field. Application inputs, outputs, metadata, and JSON Schemas can contain arbitrary JSON. ## list\_page\_views [#list_page_views] List project page views with bounded pagination. Follow nextCursor. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `cursor` | string | No | | `limit` | integer | No | | `search` | string | No | | `resource` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | ## get\_page\_view [#get_page_view] Read a page-view, including source/settings and current or pinned revision. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `id` | string | Yes | | `revision` | integer | No | **Result** Type: **object**. ## create\_page\_view [#create_page_view] Create a shared page-view. Page Views reference existing fields and never edit their definitions. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------ | ------ | -------- | | `definition` | object | Yes | **Result** Type: **object**. ## update\_page\_view [#update_page_view] Update a shared page-view using expectedRevision. Stale writes return CONFLICT. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------ | ------ | -------- | | `id` | string | Yes | | `definition` | object | Yes | **Result** Type: **object**. ## copy\_page\_view [#copy_page_view] Copy a page-view under a new name. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------ | ------ | -------- | | `id` | string | Yes | | `name` | string | Yes | **Result** Type: **object**. ## delete\_page\_view [#delete_page_view] Delete a page-view at expectedRevision. Referenced definitions must be detached first. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `expectedRevision` | integer | Yes | **Result** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | ## get\_page\_view\_history [#get_page_view_history] Read immutable page-view revisions. before paginates older revisions. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | -------- | ------- | -------- | | `id` | string | Yes | | `before` | integer | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or number | Yes | ## restore\_page\_view [#restore_page_view] Restore a historical page-view as a new revision without modifying its dependencies. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `revision` | integer | Yes | | `expectedRevision` | integer | Yes | **Result** Type: **object**. ## get\_page\_view\_dependencies [#get_page_view_dependencies] List definitions referencing this page-view; hasMore reports truncation. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** Type: **object**. ## validate\_page\_view [#validate_page_view] Validate a page-view and its dependencies. This does not execute formulas or render React. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ------------ | ------ | -------- | | `definition` | object | Yes | **Result** | Field | Type | Required | | ----------- | ------- | -------- | | `evaluated` | boolean | Yes | | `rendered` | boolean | Yes | | `valid` | boolean | Yes | ## list\_custom\_fields [#list_custom_fields] List project custom fields with bounded pagination. Follow nextCursor. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `cursor` | string | No | | `limit` | integer | No | | `search` | string | No | | `resource` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | ## get\_custom\_field [#get_custom_field] Read a custom-field, including source/settings and current or pinned revision. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `id` | string | Yes | | `revision` | integer | No | **Result** Type: **object**. ## create\_custom\_field [#create_custom_field] Create a shared custom-field. Page Views reference existing fields and never edit their definitions. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------ | ------ | -------- | | `definition` | object | Yes | **Result** Type: **object**. ## update\_custom\_field [#update_custom_field] Update a shared custom-field using expectedRevision. Stale writes return CONFLICT. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------ | ------ | -------- | | `id` | string | Yes | | `definition` | object | Yes | **Result** Type: **object**. ## copy\_custom\_field [#copy_custom_field] Copy a custom-field under a new name. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------ | ------ | -------- | | `id` | string | Yes | | `name` | string | Yes | **Result** Type: **object**. ## delete\_custom\_field [#delete_custom_field] Delete a custom-field at expectedRevision. Referenced definitions must be detached first. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `expectedRevision` | integer | Yes | **Result** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | ## get\_custom\_field\_history [#get_custom_field_history] Read immutable custom-field revisions. before paginates older revisions. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | -------- | ------- | -------- | | `id` | string | Yes | | `before` | integer | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or number | Yes | ## restore\_custom\_field [#restore_custom_field] Restore a historical custom-field as a new revision without modifying its dependencies. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `revision` | integer | Yes | | `expectedRevision` | integer | Yes | **Result** Type: **object**. ## get\_custom\_field\_dependencies [#get_custom_field_dependencies] List definitions referencing this custom-field; hasMore reports truncation. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** Type: **object**. ## validate\_custom\_field [#validate_custom_field] Validate a custom-field and its dependencies. This does not execute formulas or render React. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ------------ | ------ | -------- | | `definition` | object | Yes | **Result** | Field | Type | Required | | ----------- | ------- | -------- | | `evaluated` | boolean | Yes | | `rendered` | boolean | Yes | | `valid` | boolean | Yes | ## list\_object\_views [#list_object_views] List project object views with bounded pagination. Follow nextCursor. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `cursor` | string | No | | `limit` | integer | No | | `search` | string | No | | `resource` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | ## get\_object\_view [#get_object_view] Read a object-view, including source/settings and current or pinned revision. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `id` | string | Yes | | `revision` | integer | No | **Result** Type: **object**. ## create\_object\_view [#create_object_view] Create a shared object-view. Page Views reference existing fields and never edit their definitions. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------ | ------ | -------- | | `definition` | object | Yes | **Result** Type: **object**. ## update\_object\_view [#update_object_view] Update a shared object-view using expectedRevision. Stale writes return CONFLICT. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------ | ------ | -------- | | `id` | string | Yes | | `definition` | object | Yes | **Result** Type: **object**. ## copy\_object\_view [#copy_object_view] Copy a object-view under a new name. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------ | ------ | -------- | | `id` | string | Yes | | `name` | string | Yes | **Result** Type: **object**. ## delete\_object\_view [#delete_object_view] Delete a object-view at expectedRevision. Referenced definitions must be detached first. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `expectedRevision` | integer | Yes | **Result** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | ## get\_object\_view\_history [#get_object_view_history] Read immutable object-view revisions. before paginates older revisions. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | -------- | ------- | -------- | | `id` | string | Yes | | `before` | integer | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or number | Yes | ## restore\_object\_view [#restore_object_view] Restore a historical object-view as a new revision without modifying its dependencies. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `revision` | integer | Yes | | `expectedRevision` | integer | Yes | **Result** Type: **object**. ## get\_object\_view\_dependencies [#get_object_view_dependencies] List definitions referencing this object-view; hasMore reports truncation. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** Type: **object**. ## validate\_object\_view [#validate_object_view] Validate a object-view and its dependencies. This does not execute formulas or render React. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ------------ | ------ | -------- | | `definition` | object | Yes | **Result** | Field | Type | Required | | ----------- | ------- | -------- | | `evaluated` | boolean | Yes | | `rendered` | boolean | Yes | | `valid` | boolean | Yes | ## get\_view\_capabilities [#get_view_capabilities] Discover Page View resources, Custom Field object types, execution limits and supported operations. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** | Field | Type | Required | | ------------------- | ------------- | -------- | | `customFieldQuery` | object | Yes | | `objectTypes` | array of JSON | Yes | | `objectViewPreview` | object | Yes | | `preferences` | object | Yes | | `resources` | object | Yes | ## resolve\_page\_view [#resolve_page_view] Resolve a Page View, its exact field/Object View dependencies and workspace-relative link before applying it. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `id` | string | Yes | | `revision` | integer | No | **Result** | Field | Type | Required | | ------------- | --------------- | -------- | | `applied` | boolean | Yes | | `fields` | array of object | Yes | | `objectViews` | object | Yes | | `path` | string | Yes | | `view` | object | Yes | ## get\_page\_view\_data [#get_page_view_data] Query a Page View with data permissions. Unsupported query operations are explicitly rejected; rows are never filtered after pagination. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | -------- | ------- | -------- | | `id` | string | Yes | | `limit` | integer | No | | `offset` | integer | No | | `runId` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `columns` | array of object | Yes | | `limit` | number | Yes | | `nextOffset` | null or number | Yes | | `offset` | number | Yes | | `rows` | array of object | Yes | | `total` | number | Yes | | `view` | object | Yes | ## evaluate\_custom\_field [#evaluate_custom_field] Evaluate up to 20 supplied rows in the bounded local sandbox, returning typed values and per-row diagnostics. Does not fetch records or call model providers. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ---------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `revision` | integer | No | | `kind` | "trace" or "dataset-item" or "session" or "eval-run" or "eval-result" or "dataset" or "app" or "agent" or "workflow" or "review-session" or "review-item" or "prompt" or "scorer" or "human-score" or "score-collection" or "dashboard" or "alert" or "notification" or "table-row" | Yes | | `rows` | array of object | Yes | **Result** | Field | Type | Required | | ----------- | --------------- | -------- | | `evaluated` | boolean | Yes | | `id` | string | Yes | | `limits` | object | Yes | | `query` | object | Yes | | `results` | array of object | Yes | | `revision` | number | Yes | ## preview\_object\_view [#preview_object_view] Compile an Object View and check input compatibility. Returns rendered:false; use the browser sandbox for actual rendering. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ---------- | ------------------------------------------------------------------------------------------------------------------------ | -------- | | `id` | string | Yes | | `revision` | integer | No | | `kind` | "trace" or "dataset-item" | Yes | | `object` | string or number or boolean or null or array of string or number or boolean or null or array of JSON or object or object | Yes | | `context` | object | No | **Result** Type: **object**. ## get\_view\_preference [#get_view_preference] Read the authenticated principal's durable project/page preference. Revision zero means unset. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ------- | ------ | -------- | | `scope` | string | Yes | **Result** | Field | Type | Required | | ---------- | ------ | -------- | | `revision` | number | Yes | | `scope` | string | Yes | | `value` | JSON | Yes | ## save\_view\_preference [#save_view_preference] Save the authenticated principal's durable project/page preference using expectedRevision. Does not edit shared definitions. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `scope` | string | Yes | | `expectedRevision` | integer | Yes | | `value` | object | Yes | **Result** | Field | Type | Required | | ---------- | ------ | -------- | | `revision` | number | Yes | | `scope` | string | Yes | | `value` | object | Yes | ## list\_apps [#list_apps] List registered apps, connection types, input/output schemas and availability. HTTP apps are always available; availability is not a health check. Required scopes: `apps:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** — array; each item has these fields | Field | Type | Required | | ----------------- | ------------------ | -------- | | `codeProvenance` | JSON | No | | `connection` | object | Yes | | `defaultInput` | JSON | No | | `evaluatorIds` | array of string | Yes | | `id` | string | Yes | | `inputSchema` | object | Yes | | `internalTracing` | boolean | Yes | | `mode` | "input" or "agent" | Yes | | `name` | string | Yes | | `online` | boolean | Yes | | `outputSchema` | object | Yes | | `revision` | number | Yes | ## get\_app [#get_app] Read one app and its connection configuration. Credential values are never returned. Required scopes: `apps:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ----------------- | ------------------ | -------- | | `codeProvenance` | JSON | No | | `connection` | object | Yes | | `defaultInput` | JSON | No | | `evaluatorIds` | array of string | Yes | | `id` | string | Yes | | `inputSchema` | object | Yes | | `internalTracing` | boolean | Yes | | `mode` | "input" or "agent" | Yes | | `name` | string | Yes | | `online` | boolean | Yes | | `outputSchema` | object | Yes | | `revision` | number | Yes | ## register\_app [#register_app] Register an app with a local bridge or HTTP webhook connection. Use expectedRevision when editing. Omitted webhook headers preserve saved credentials; \{} clears them. Registration does not call the app. Required scopes: `apps:write`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `app` | object | Yes | **Result** | Field | Type | Required | | ----------------- | ------------------ | -------- | | `codeProvenance` | JSON | No | | `connection` | object | Yes | | `defaultInput` | JSON | No | | `evaluatorIds` | array of string | Yes | | `id` | string | Yes | | `inputSchema` | object | Yes | | `internalTracing` | boolean | Yes | | `mode` | "input" or "agent" | Yes | | `name` | string | Yes | | `outputSchema` | object | Yes | | `revision` | number | Yes | ## run\_app [#run_app] Run one input through a registered app and save its experiment and traces. Reuse requestKey after uncertain delivery to prevent duplicate execution; inspect eval runs if the request is still starting. scorerIds omitted uses app defaults; \[] runs without scorers. Calls may incur cost. Required scopes: `apps:read`, `apps:write`, `evals:read`, `evals:write`, `scorers:read`. **Request fields** | Field | Type | Required | | ------------ | ------------------------------------------------------------------------------------------------------------------------ | -------- | | `id` | string | Yes | | `input` | string or number or boolean or null or array of string or number or boolean or null or array of JSON or object or object | Yes | | `requestKey` | string | Yes | | `scorerIds` | array of string | No | **Result** | Field | Type | Required | | --------------------- | ---------------------------------------------------------------- | -------- | | `completedAt` | null or string | Yes | | `createdAt` | string | Yes | | `datasetId` | null or string | Yes | | `evaluatorIds` | array of string | Yes | | `evaluatorVersionIds` | JSON | No | | `execution` | JSON | No | | `groups` | array of object | No | | `groupsResolvedAt` | null or string | No | | `id` | string | Yes | | `metadata` | object | Yes | | `name` | null or string | Yes | | `nextCursor` | null or string | No | | `resultCount` | number | Yes | | `results` | array of object | Yes | | `rows` | array of object | No | | `score` | null or number | Yes | | `scorerProgress` | array of object | No | | `scores` | array of object | No | | `status` | "cancelled" or "completed" or "failed" or "partial" or "running" | Yes | | `targetCount` | number | Yes | | `traceCount` | number | Yes | | `url` | string | Yes | ## get\_metrics\_metadata [#get_metrics_metadata] Discover semantic sources, measures, dimensions, time dimensions, segments, and query limits before constructing metric queries. Prefer source.visibility=primary for new queries: Traces (traces), Spans (spans), Evaluation Runs (evalRuns), Evaluation Results (evalResults), Scores (scoreValues). Inspect member definitions, denominators, grouping/filter capabilities and source.unavailable; legacy scores is not the Scores ratings source. Required scopes: `metrics:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** | Field | Type | Required | | -------- | --------------- | -------- | | `models` | array of object | Yes | ## query\_metrics [#query_metrics] Run a bounded read-only semantic query using discovered catalog members and the source's time dimension. Score measures marked requiresDefinition need one compatible scoreValues.definitionId or grouping by it. Returns data, annotations, quality, and execution metadata; does not accept SQL or executable definitions. On READ\_BUSY, honor details.retryAfterSeconds and retry this read with bounded backoff. Required scopes: `metrics:read`. **Request fields** | Field | Type | Required | | ------- | ------ | -------- | | `query` | object | Yes | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `annotation` | object | Yes | | `data` | array of object | Yes | | `meta` | object | Yes | | `query` | object | Yes | ## batch\_metrics [#batch_metrics] Run 1–40 read-only semantic queries sequentially against one consistent snapshot. Returns results in query order with shared snapshot metadata. Use discovered catalog members and source-specific time dimensions; on READ\_BUSY, honor details.retryAfterSeconds and retry this read with bounded backoff. Required scopes: `metrics:read`. **Request fields** | Field | Type | Required | | --------- | --------------- | -------- | | `queries` | array of object | Yes | **Result** — array; each item has these fields | Field | Type | Required | | ------------ | --------------- | -------- | | `annotation` | object | Yes | | `data` | array of object | Yes | | `meta` | object | Yes | | `query` | object | Yes | ## list\_human\_scores [#list_human_scores] Read the project Human Score library and reusable collections. Human Scores are review criteria, separate from automated Scorers. Required scopes: `reviews:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** | Field | Type | Required | | ------------- | --------------- | -------- | | `collections` | array of object | Yes | | `scores` | array of object | Yes | ## get\_human\_score [#get_human_score] Read one Human Score definition, including type, numeric range or enum options and revision. Required scopes: `reviews:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** Type: **object**. ## create\_human\_score [#create_human_score] Create a reusable numeric, categorical (single or multiple choice), or text Human Score in this project. Required scopes: `reviews:write`. **Request fields** | Field | Type | Required | | ------- | ------ | -------- | | `score` | object | Yes | **Result** Type: **object**. ## update\_human\_score [#update_human_score] Update a Human Score with its current revision. Existing session and rating snapshots retain their definitions. Required scopes: `reviews:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `expectedRevision` | integer | Yes | | `score` | object | Yes | **Result** Type: **object**. ## get\_human\_score\_collection [#get_human_score_collection] Read an ordered collection with full Human Score definitions. Required scopes: `reviews:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ------------- | --------------- | -------- | | `archived` | boolean | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `revision` | number | Yes | | `scoreIds` | array of string | Yes | | `scores` | array of object | Yes | | `updatedAt` | string | Yes | ## create\_human\_score\_collection [#create_human_score_collection] Create an ordered reusable collection of project Human Score IDs. Attach it to a review session using collectionId. Required scopes: `reviews:write`. **Request fields** | Field | Type | Required | | ------------- | --------------- | -------- | | `name` | string | Yes | | `description` | string | No | | `scoreIds` | array of string | Yes | **Result** | Field | Type | Required | | ------------- | --------------- | -------- | | `archived` | boolean | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `revision` | number | Yes | | `scoreIds` | array of string | Yes | | `scores` | array of object | Yes | | `updatedAt` | string | Yes | ## update\_human\_score\_collection [#update_human_score_collection] Update an ordered Human Score collection. Attached sessions retain their snapshot. Required scopes: `reviews:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `expectedRevision` | integer | Yes | | `collection` | object | Yes | **Result** | Field | Type | Required | | ------------- | --------------- | -------- | | `archived` | boolean | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `revision` | number | Yes | | `scoreIds` | array of string | Yes | | `scores` | array of object | Yes | | `updatedAt` | string | Yes | ## list\_review\_sessions [#list_review_sessions] List project review sessions with assignment, explicit AI labels, and separate human/AI completion counts. reviewedCount is total score completion, not human verification. Filters include name, status, and assigneeName. Required scopes: `reviews:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `filter` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## get\_review\_session [#get_review_session] Read a review session, its review prompt, and up to 500 trace items in their fixed playback order. Required scopes: `reviews:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | -------------------- | ------------------------------------------------- | -------- | | `aiLabelledCount` | number | No | | `aiReviewedCount` | number | No | | `assigneeName` | null or string | Yes | | `assigneeUserId` | null or string | Yes | | `collection` | null or object | Yes | | `collectionId` | null or string | Yes | | `createdAt` | string | Yes | | `createdBy` | null or string | Yes | | `humanReviewedCount` | number | No | | `id` | string | Yes | | `items` | array of object | Yes | | `label` | "AI-labelled" or "Human-reviewed" or "Unreviewed" | No | | `name` | string | Yes | | `number` | number | Yes | | `prompt` | string | Yes | | `reviewedCount` | number | Yes | | `reviewers` | array of object | Yes | | `revision` | number | Yes | | `skippedCount` | number | Yes | | `status` | "completed" or "pending" or "in\_progress" | Yes | | `traceCount` | number | Yes | | `updatedAt` | string | Yes | ## get\_review\_options [#get_review_options] List members who can be assigned to this project and the current reviewer identity. Use list\_human\_scores for the separate Human Score library. Required scopes: `reviews:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** | Field | Type | Required | | --------------- | --------------- | -------- | | `currentUserId` | null or string | Yes | | `members` | array of object | Yes | ## create\_review\_session [#create_review_session] Create an ordered collection of traces for review. Use list\_traces to find traces for a prompt, then pass their IDs in the desired order. The prompt field supplies review instructions. Optional collectionId attaches a snapshot of a Human Score collection to every trace. Does not run a model or execute the traces. Required scopes: `reviews:write`, `traces:read`. **Request fields** | Field | Type | Required | | ----------------- | --------------- | -------- | | `idempotencyKey` | string | No | | `name` | string | No | | `prompt` | string | No | | `assigneeUserId` | string or null | No | | `reviewerUserIds` | array of string | No | | `collectionId` | string or null | No | | `traceIds` | array of string | Yes | **Result** | Field | Type | Required | | -------------------- | ------------------------------------------------- | -------- | | `aiLabelledCount` | number | No | | `aiReviewedCount` | number | No | | `assigneeName` | null or string | Yes | | `assigneeUserId` | null or string | Yes | | `collection` | null or object | Yes | | `collectionId` | null or string | Yes | | `createdAt` | string | Yes | | `createdBy` | null or string | Yes | | `humanReviewedCount` | number | No | | `id` | string | Yes | | `items` | array of object | Yes | | `label` | "AI-labelled" or "Human-reviewed" or "Unreviewed" | No | | `name` | string | Yes | | `number` | number | Yes | | `prompt` | string | Yes | | `reviewedCount` | number | Yes | | `reviewers` | array of object | Yes | | `revision` | number | Yes | | `skippedCount` | number | Yes | | `status` | "completed" or "pending" or "in\_progress" | Yes | | `traceCount` | number | Yes | | `updatedAt` | string | Yes | ## update\_review\_session [#update_review_session] Edit the review prompt, name, assigned project members, or collection using the current revision. Collection changes recalculate completion while retaining existing ratings and their provenance. Required scopes: `reviews:write`. **Request fields** | Field | Type | Required | | ------------------ | --------------- | -------- | | `id` | string | Yes | | `expectedRevision` | integer | Yes | | `collectionId` | string or null | No | | `name` | string | No | | `prompt` | string | No | | `assigneeUserId` | string or null | No | | `reviewerUserIds` | array of string | No | **Result** | Field | Type | Required | | -------------------- | ------------------------------------------------- | -------- | | `aiLabelledCount` | number | No | | `aiReviewedCount` | number | No | | `assigneeName` | null or string | Yes | | `assigneeUserId` | null or string | Yes | | `collection` | null or object | Yes | | `collectionId` | null or string | Yes | | `createdAt` | string | Yes | | `createdBy` | null or string | Yes | | `humanReviewedCount` | number | No | | `id` | string | Yes | | `items` | array of object | Yes | | `label` | "AI-labelled" or "Human-reviewed" or "Unreviewed" | No | | `name` | string | Yes | | `number` | number | Yes | | `prompt` | string | Yes | | `reviewedCount` | number | Yes | | `reviewers` | array of object | Yes | | `revision` | number | Yes | | `skippedCount` | number | Yes | | `status` | "completed" or "pending" or "in\_progress" | Yes | | `traceCount` | number | Yes | | `updatedAt` | string | Yes | ## get\_review\_item [#get_review_item] Read one review item's scores, notes, revision, trace ID and previous/next IDs. Use get\_trace to inspect the actual prompt and output before reviewing. Required scopes: `reviews:read`. **Request fields** | Field | Type | Required | | ----------- | ------ | -------- | | `sessionId` | string | Yes | | `itemId` | string | Yes | **Result** | Field | Type | Required | | ----------------- | ------------------------------------------------------------------------- | -------- | | `annotations` | array of object | Yes | | `completionKind` | null or "ai" or "human" | No | | `definitions` | array of object | Yes | | `humanVerified` | false or true | No | | `id` | string | Yes | | `label` | "AI-labelled" or "Human-reviewed" or "Unreviewed" or "Unknown provenance" | No | | `lastSubmission` | null or object | No | | `nextItemId` | null or string | Yes | | `notes` | string | Yes | | `notesProvenance` | null or object | No | | `ordinal` | number | Yes | | `previousItemId` | null or string | Yes | | `reviewedAt` | null or string | Yes | | `reviewedBy` | null or string | Yes | | `revision` | number | Yes | | `scores` | array of object | Yes | | `sessionId` | string | Yes | | `skippedAt` | null or string | Yes | | `traceId` | string | Yes | | `traceName` | string | Yes | ## export\_review\_items [#export_review_items] Export a bounded page of review items with complete scores, notes, annotations and trusted provenance. AI-labelled rows are not human-verified ground truth. Follow nextCursor to export the session. Required scopes: `reviews:read`. **Request fields** | Field | Type | Required | | -------- | -------------- | -------- | | `id` | string | Yes | | `cursor` | string or null | No | | `limit` | integer | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | ## record\_review [#record_review] Save AI-labelled scores, notes or annotations for one trace using an organization API key or user OAuth with reviews:write. Attribution is derived server-side from the authenticated principal; optional agent \{name, model} is descriptive metadata only. Providing scores replaces the complete score set; include every required criterion to complete it. Omit scores for notes/annotations-only updates that preserve scores and completion. Use humanScoreId and humanScoreRevision from get\_review\_item or list\_human\_scores; null is an unanswered draft. Values must follow the rubric's numeric range, option values, multiple-choice array or text type. Requires expectedRevision from get\_review\_item; on conflict refetch before retrying. AI completion is separate from human verification and never updates dataset ground truth. Returns provenance, completionKind, humanVerified and nextItemId. Required scopes: `reviews:write`. **Request fields** | Field | Type | Required | | ------------------ | --------------- | -------- | | `sessionId` | string | Yes | | `itemId` | string | Yes | | `expectedRevision` | integer | Yes | | `scores` | array of object | No | | `notes` | string | No | | `annotations` | array of object | No | | `agent` | object | No | **Result** | Field | Type | Required | | ----------------- | ------------------------------------------------------------------------- | -------- | | `annotations` | array of object | Yes | | `completionKind` | null or "ai" or "human" | No | | `definitions` | array of object | Yes | | `humanVerified` | false or true | No | | `id` | string | Yes | | `label` | "AI-labelled" or "Human-reviewed" or "Unreviewed" or "Unknown provenance" | No | | `lastSubmission` | null or object | No | | `nextItemId` | null or string | Yes | | `notes` | string | Yes | | `notesProvenance` | null or object | No | | `ordinal` | number | Yes | | `previousItemId` | null or string | Yes | | `reviewedAt` | null or string | Yes | | `reviewedBy` | null or string | Yes | | `revision` | number | Yes | | `scores` | array of object | Yes | | `sessionId` | string | Yes | | `skippedAt` | null or string | Yes | | `traceId` | string | Yes | | `traceName` | string | Yes | ## list\_scorer\_libraries [#list_scorer_libraries] Discover supported evaluator libraries, pinned package/adapter versions, required arguments and model requirements. Create a type=library scorer with library \{package, version, adapterVersion, evaluator, mappings, options}. Input mappings use trace or datasetItem dotted paths. Factuality uses the project's Vercel AI Gateway language model; deterministic evaluators need no provider. Required scopes: `scorers:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** | Field | Type | Required | | ----------------- | --------------- | -------- | | `defaultMappings` | object | Yes | | `libraries` | array of object | Yes | ## use\_library\_scorer [#use_library_scorer] Select a library evaluator using its default input mappings. Creates or reuses its saved project scorer without overwriting edits. Factuality defaults to openai/gpt-4.1-mini via the project's Gateway. Use the returned ID anywhere scorer IDs are accepted. Required scopes: `scorers:write`. **Request fields** | Field | Type | Required | | ----------- | ------------------------------------------------------------------------------------------- | -------- | | `evaluator` | "ExactMatch" or "Levenshtein" or "NumericDiff" or "ValidJSON" or "JSONDiff" or "Factuality" | Yes | **Result** | Field | Type | Required | | ---------------- | ------------------------------------------------------------ | -------- | | `allowSkip` | boolean | Yes | | `chainOfThought` | boolean | Yes | | `choices` | array of object | Yes | | `code` | string | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `id` | string | Yes | | `imagePaths` | array of string | No | | `library` | JSON | No | | `messages` | array of object | Yes | | `model` | string | Yes | | `modelType` | "language" or "evaluation" | No | | `name` | string | Yes | | `provider` | "datool" or "vercel-ai-gateway" or "openai" or "typesafe-ai" | No | | `revision` | number | Yes | | `slug` | string | Yes | | `threshold` | null or number | Yes | | `type` | "llm" or "javascript" or "python" or "library" | Yes | | `updatedAt` | string | Yes | ## list\_scorers [#list_scorers] List saved LLM, JavaScript, Python and library scorer configurations. Required scopes: `scorers:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** — array; each item has these fields | Field | Type | Required | | ---------------- | ------------------------------------------------------------ | -------- | | `allowSkip` | boolean | Yes | | `chainOfThought` | boolean | Yes | | `choices` | array of object | Yes | | `code` | string | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `id` | string | Yes | | `imagePaths` | array of string | No | | `library` | JSON | No | | `messages` | array of object | Yes | | `model` | string | Yes | | `modelType` | "language" or "evaluation" | No | | `name` | string | Yes | | `provider` | "datool" or "vercel-ai-gateway" or "openai" or "typesafe-ai" | No | | `revision` | number | Yes | | `slug` | string | Yes | | `threshold` | null or number | Yes | | `type` | "llm" or "javascript" or "python" or "library" | Yes | | `updatedAt` | string | Yes | ## get\_scorer [#get_scorer] Read a scorer by ID. Required scopes: `scorers:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ---------------- | ------------------------------------------------------------ | -------- | | `allowSkip` | boolean | Yes | | `chainOfThought` | boolean | Yes | | `choices` | array of object | Yes | | `code` | string | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `id` | string | Yes | | `imagePaths` | array of string | No | | `library` | JSON | No | | `messages` | array of object | Yes | | `model` | string | Yes | | `modelType` | "language" or "evaluation" | No | | `name` | string | Yes | | `provider` | "datool" or "vercel-ai-gateway" or "openai" or "typesafe-ai" | No | | `revision` | number | Yes | | `slug` | string | Yes | | `threshold` | null or number | Yes | | `type` | "llm" or "javascript" or "python" or "library" | Yes | | `updatedAt` | string | Yes | ## create\_scorer [#create_scorer] Create a scorer configuration; does not execute code or call a model. Required scopes: `scorers:write`. **Request fields** | Field | Type | Required | | -------- | ------ | -------- | | `scorer` | object | Yes | **Result** | Field | Type | Required | | ---------------- | ------------------------------------------------------------ | -------- | | `allowSkip` | boolean | Yes | | `chainOfThought` | boolean | Yes | | `choices` | array of object | Yes | | `code` | string | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `id` | string | Yes | | `imagePaths` | array of string | No | | `library` | JSON | No | | `messages` | array of object | Yes | | `model` | string | Yes | | `modelType` | "language" or "evaluation" | No | | `name` | string | Yes | | `provider` | "datool" or "vercel-ai-gateway" or "openai" or "typesafe-ai" | No | | `revision` | number | Yes | | `slug` | string | Yes | | `threshold` | null or number | Yes | | `type` | "llm" or "javascript" or "python" or "library" | Yes | | `updatedAt` | string | Yes | ## update\_scorer [#update_scorer] Replace the complete scorer configuration by ID. Required scopes: `scorers:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `scorer` | object | Yes | | `expectedRevision` | integer | Yes | **Result** | Field | Type | Required | | ---------------- | ------------------------------------------------------------ | -------- | | `allowSkip` | boolean | Yes | | `chainOfThought` | boolean | Yes | | `choices` | array of object | Yes | | `code` | string | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `id` | string | Yes | | `imagePaths` | array of string | No | | `library` | JSON | No | | `messages` | array of object | Yes | | `model` | string | Yes | | `modelType` | "language" or "evaluation" | No | | `name` | string | Yes | | `provider` | "datool" or "vercel-ai-gateway" or "openai" or "typesafe-ai" | No | | `revision` | number | Yes | | `slug` | string | Yes | | `threshold` | null or number | Yes | | `type` | "llm" or "javascript" or "python" or "library" | Yes | | `updatedAt` | string | Yes | ## delete\_scorer [#delete_scorer] Permanently delete a scorer configuration. Required scopes: `scorers:write`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | ## list\_views [#list_views] List custom eval-run table views. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** — array; each item has these fields | Field | Type | Required | | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- | | `createdAt` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `resource` | "traces" or "alerts" or "reviews" or "prompts" or "reports" or "scorers" or "agents" or "workflows" or "sessions" or "eval-runs" or "review-session" or "session-traces" or "evaluations" or "eval-results" or "eval-comparison" or "datasets" or "dataset-items" or "dataset-item-runs" or "playground-apps" or "playground-traces" or "human-scores" or "score-collections" or "dashboards" or "dashboard-table" or "notifications" | Yes | | `revision` | number | Yes | | `settings` | object | Yes | | `updatedAt` | string | Yes | ## get\_view [#get_view] Read a custom eval-run view including its revision. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- | | `createdAt` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `resource` | "traces" or "alerts" or "reviews" or "prompts" or "reports" or "scorers" or "agents" or "workflows" or "sessions" or "eval-runs" or "review-session" or "session-traces" or "evaluations" or "eval-results" or "eval-comparison" or "datasets" or "dataset-items" or "dataset-item-runs" or "playground-apps" or "playground-traces" or "human-scores" or "score-collections" or "dashboards" or "dashboard-table" or "notifications" | Yes | | `revision` | number | Yes | | `settings` | object | Yes | | `updatedAt` | string | Yes | ## create\_view [#create_view] Create a custom eval-run view. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ---------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- | | `resource` | "traces" or "sessions" or "session-traces" or "eval-runs" or "evaluations" or "eval-results" or "eval-comparison" or "datasets" or "dataset-items" or "dataset-item-runs" or "playground-apps" or "playground-traces" or "agents" or "workflows" or "reviews" or "review-session" or "prompts" or "scorers" or "human-scores" or "score-collections" or "dashboards" or "reports" or "dashboard-table" or "alerts" or "notifications" | Yes | | `name` | string | Yes | | `settings` | object | Yes | **Result** | Field | Type | Required | | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- | | `createdAt` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `resource` | "traces" or "alerts" or "reviews" or "prompts" or "reports" or "scorers" or "agents" or "workflows" or "sessions" or "eval-runs" or "review-session" or "session-traces" or "evaluations" or "eval-results" or "eval-comparison" or "datasets" or "dataset-items" or "dataset-item-runs" or "playground-apps" or "playground-traces" or "human-scores" or "score-collections" or "dashboards" or "dashboard-table" or "notifications" | Yes | | `revision` | number | Yes | | `settings` | object | Yes | | `updatedAt` | string | Yes | ## update\_view [#update_view] Replace a custom view with revision conflict protection. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- | | `resource` | "traces" or "sessions" or "session-traces" or "eval-runs" or "evaluations" or "eval-results" or "eval-comparison" or "datasets" or "dataset-items" or "dataset-item-runs" or "playground-apps" or "playground-traces" or "agents" or "workflows" or "reviews" or "review-session" or "prompts" or "scorers" or "human-scores" or "score-collections" or "dashboards" or "reports" or "dashboard-table" or "alerts" or "notifications" | Yes | | `name` | string | Yes | | `settings` | object | Yes | | `expectedRevision` | integer | Yes | | `id` | string | Yes | **Result** | Field | Type | Required | | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- | | `createdAt` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `resource` | "traces" or "alerts" or "reviews" or "prompts" or "reports" or "scorers" or "agents" or "workflows" or "sessions" or "eval-runs" or "review-session" or "session-traces" or "evaluations" or "eval-results" or "eval-comparison" or "datasets" or "dataset-items" or "dataset-item-runs" or "playground-apps" or "playground-traces" or "human-scores" or "score-collections" or "dashboards" or "dashboard-table" or "notifications" | Yes | | `revision` | number | Yes | | `settings` | object | Yes | | `updatedAt` | string | Yes | ## delete\_view [#delete_view] Permanently delete a custom view at its expected revision. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `expectedRevision` | integer | Yes | **Result** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | ## list\_saved\_views [#list_saved_views] List saved trace and eval-result selector views (distinct from custom eval-run views). Required scopes: `views:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## get\_saved\_view [#get_saved_view] Read a saved selector view. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------- | -------- | | `columns` | array of object | Yes | | `createdAt` | string | Yes | | `filters` | array of object | Yes | | `id` | string | Yes | | `name` | string | Yes | | `resource` | "traces" or "eval-results" | Yes | | `sort` | null or object | Yes | | `updatedAt` | string | Yes | ## create\_saved\_view [#create_saved_view] Create a saved trace or eval-result selector view. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ---------- | -------------------------- | -------- | | `columns` | array of object | Yes | | `filters` | array of object | No | | `id` | string | No | | `name` | string | Yes | | `resource` | "eval-results" or "traces" | Yes | | `sort` | object or null | No | **Result** | Field | Type | Required | | ----------- | -------------------------- | -------- | | `columns` | array of object | Yes | | `createdAt` | string | Yes | | `filters` | array of object | Yes | | `id` | string | Yes | | `name` | string | Yes | | `resource` | "traces" or "eval-results" | Yes | | `sort` | null or object | Yes | | `updatedAt` | string | Yes | ## update\_saved\_view [#update_saved_view] Patch a saved selector view. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ------- | ------ | -------- | | `id` | string | Yes | | `patch` | object | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------- | -------- | | `columns` | array of object | Yes | | `createdAt` | string | Yes | | `filters` | array of object | Yes | | `id` | string | Yes | | `name` | string | Yes | | `resource` | "traces" or "eval-results" | Yes | | `sort` | null or object | Yes | | `updatedAt` | string | Yes | ## delete\_saved\_view [#delete_saved_view] Permanently delete a saved selector view. Required scopes: `views:write`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | ## list\_datasets [#list_datasets] List datasets with cursor pagination. Required scopes: `datasets:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `filter` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## get\_dataset [#get_dataset] Read a dataset summary and its first item page. Set includeItems=false for header/count only. Use list\_dataset\_items for more. Required scopes: `datasets:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `id` | string | Yes | | `includeItems` | boolean | No | **Result** | Field | Type | Required | | -------------- | --------------- | -------- | | `createdAt` | string | Yes | | `description` | null or string | Yes | | `fieldSchemas` | JSON | No | | `id` | string | Yes | | `itemCount` | number | Yes | | `items` | array of object | Yes | | `metadata` | JSON | No | | `name` | string | Yes | | `nextCursor` | null or string | No | | `revision` | number | No | | `updatedAt` | string | Yes | | `versionId` | string | No | ## list\_dataset\_items [#list_dataset_items] Read a bounded page of dataset items. Set preview=true to omit large values with explicit omittedFields markers; fetch complete values with get\_dataset\_item before editing or exporting. Required scopes: `datasets:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `id` | string | Yes | | `preview` | boolean | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## get\_dataset\_item [#get_dataset_item] Read one dataset item. Optionally set fields to load only those fields in full, keeping other large fields as previews. Omit fields for the complete item. Capped at 32 MiB. Required scopes: `datasets:read`. **Request fields** | Field | Type | Required | | -------- | -------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `fields` | array of "input" or "expectedOutput" or "metadata" or "sourceSpanEvidence" | No | **Result** | Field | Type | Required | | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------ | -------- | | `createdAt` | string | Yes | | `datasetId` | string | Yes | | `datasetRevision` | number | No | | `datasetVersionId` | string | No | | `expectedOutput` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `id` | string | Yes | | `input` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `metadata` | object | Yes | | `observedOutput` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | No | | `omittedFields` | JSON | No | | `sourceSpanEvidence` | null or object | No | | `sourceSpanId` | null or string | No | | `sourceTraceId` | null or string | Yes | | `updatedAt` | string | Yes | | `versionId` | string | No | ## create\_dataset [#create_dataset] Create a dataset. Required scopes: `datasets:write`. **Request fields** | Field | Type | Required | | ------------- | ------ | -------- | | `description` | string | No | | `id` | string | No | | `name` | string | Yes | **Result** | Field | Type | Required | | -------------- | -------------- | -------- | | `createdAt` | string | Yes | | `description` | null or string | Yes | | `fieldSchemas` | JSON | No | | `id` | string | Yes | | `itemCount` | number | Yes | | `metadata` | JSON | No | | `name` | string | Yes | | `revision` | number | No | | `updatedAt` | string | Yes | | `versionId` | string | No | ## update\_dataset [#update_dataset] Patch dataset name or description. Required scopes: `datasets:write`. **Request fields** | Field | Type | Required | | ------- | ------ | -------- | | `id` | string | Yes | | `patch` | object | Yes | **Result** | Field | Type | Required | | -------------- | -------------- | -------- | | `createdAt` | string | Yes | | `description` | null or string | Yes | | `fieldSchemas` | JSON | No | | `id` | string | Yes | | `itemCount` | number | Yes | | `metadata` | JSON | No | | `name` | string | Yes | | `revision` | number | No | | `updatedAt` | string | Yes | | `versionId` | string | No | ## delete\_dataset [#delete_dataset] Permanently delete a dataset and its items. Datasets referenced by eval runs cannot be deleted. Required scopes: `datasets:write`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | ## create\_dataset\_item [#create_dataset_item] Add input, expected output, metadata, and an optional source trace to a dataset. Required scopes: `datasets:write`. **Request fields** | Field | Type | Required | | ----------- | ------ | -------- | | `datasetId` | string | Yes | | `item` | object | Yes | **Result** | Field | Type | Required | | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------ | -------- | | `createdAt` | string | Yes | | `datasetId` | string | Yes | | `datasetRevision` | number | No | | `datasetVersionId` | string | No | | `expectedOutput` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `id` | string | Yes | | `input` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `metadata` | object | Yes | | `observedOutput` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | No | | `sourceSpanEvidence` | null or object | No | | `sourceSpanId` | null or string | No | | `sourceTraceId` | null or string | Yes | | `updatedAt` | string | Yes | | `versionId` | string | No | ## update\_dataset\_item [#update_dataset_item] Patch a dataset item. Required scopes: `datasets:write`. **Request fields** | Field | Type | Required | | ------- | ------ | -------- | | `id` | string | Yes | | `patch` | object | Yes | **Result** | Field | Type | Required | | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------ | -------- | | `createdAt` | string | Yes | | `datasetId` | string | Yes | | `datasetRevision` | number | No | | `datasetVersionId` | string | No | | `expectedOutput` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `id` | string | Yes | | `input` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `metadata` | object | Yes | | `observedOutput` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | No | | `omittedFields` | JSON | No | | `sourceSpanEvidence` | null or object | No | | `sourceSpanId` | null or string | No | | `sourceTraceId` | null or string | Yes | | `updatedAt` | string | Yes | | `versionId` | string | No | ## delete\_dataset\_item [#delete_dataset_item] Permanently delete a dataset item. Required scopes: `datasets:write`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | ## list\_dashboards [#list_dashboards] List saved dashboard configurations. Required scopes: `dashboards:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** — array; each item has these fields | Field | Type | Required | | ------------------- | --------------- | -------- | | `createdAt` | string | Yes | | `defaultWindowDays` | number | No | | `description` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `revision` | number | Yes | | `schemaVersion` | 1 | Yes | | `updatedAt` | string | Yes | | `widgets` | array of object | Yes | ## get\_dashboard [#get_dashboard] Read a dashboard, its widgets and current revision. Required scopes: `dashboards:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ------------------- | --------------- | -------- | | `createdAt` | string | Yes | | `defaultWindowDays` | number | No | | `description` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `revision` | number | Yes | | `schemaVersion` | 1 | Yes | | `updatedAt` | string | Yes | | `widgets` | array of object | Yes | ## create\_dashboard [#create_dashboard] Create a dashboard with validated widget queries. Does not execute queries. Required scopes: `dashboards:write`. **Request fields** | Field | Type | Required | | -------- | ------ | -------- | | `config` | object | Yes | **Result** | Field | Type | Required | | ------------------- | --------------- | -------- | | `createdAt` | string | Yes | | `defaultWindowDays` | number | No | | `description` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `revision` | number | Yes | | `schemaVersion` | 1 | Yes | | `updatedAt` | string | Yes | | `widgets` | array of object | Yes | ## update\_dashboard [#update_dashboard] Replace dashboard configuration with revision conflict protection. Required scopes: `dashboards:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `config` | object | Yes | | `expectedRevision` | integer | Yes | | `id` | string | Yes | **Result** | Field | Type | Required | | ------------------- | --------------- | -------- | | `createdAt` | string | Yes | | `defaultWindowDays` | number | No | | `description` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `revision` | number | Yes | | `schemaVersion` | 1 | Yes | | `updatedAt` | string | Yes | | `widgets` | array of object | Yes | ## delete\_dashboard [#delete_dashboard] Permanently delete a dashboard at its expected revision. Required scopes: `dashboards:write`. **Request fields** | Field | Type | Required | | ------------------ | ------- | -------- | | `id` | string | Yes | | `expectedRevision` | integer | Yes | **Result** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | ## list\_traces [#list_traces] Filter and page trace summaries. Filter syntax includes status = "errored", name contains "agent", and startedAt >= "-24h". Use get\_trace for full evidence. Required scopes: `traces:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `filter` | string | No | | `sessionId` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## get\_trace [#get_trace] Get complete trace evidence including spans and scores; capped at 8 MiB. Use paged child tools for larger traces. Required scopes: `traces:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ----------------- | ------------------------------------------------------------------------------------------------------------------------------------ | -------- | | `attributes` | object | Yes | | `durationMs` | null or number | Yes | | `endedAt` | null or string | Yes | | `group` | null or object | No | | `id` | string | Yes | | `input` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `name` | string | Yes | | `nextScoreCursor` | null or string | No | | `nextSpanCursor` | null or string | No | | `operation` | string | Yes | | `output` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `scores` | array of object | Yes | | `sessionId` | null or string | Yes | | `spans` | array of object | Yes | | `spanStats` | JSON | No | | `startedAt` | string | Yes | | `status` | "cancelled" or "completed" or "running" or "errored" | Yes | ## list\_trace\_spans [#list_trace_spans] Page spans in a trace in start-time order. Required scopes: `traces:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `id` | string | Yes | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## get\_span\_path [#get_span_path] Get the ancestor path of a span within its trace. Required scopes: `traces:read`. **Request fields** | Field | Type | Required | | --------- | ------ | -------- | | `traceId` | string | Yes | | `spanId` | string | Yes | **Result** — array; each item has these fields | Field | Type | Required | | ------------ | ------------------------------------------------------------------------------------------------------------------------------------ | -------- | | `attributes` | object | Yes | | `durationMs` | null or number | Yes | | `endedAt` | null or string | Yes | | `group` | null or object | No | | `id` | string | Yes | | `input` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `kind` | "function" or "agent" or "custom" or "llm" or "score" or "workflow" or "task" or "tool" | Yes | | `name` | string | Yes | | `output` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `parentId` | null or string | Yes | | `startedAt` | string | Yes | | `status` | "cancelled" or "completed" or "running" or "errored" | Yes | | `traceId` | string | Yes | ## list\_trace\_scores [#list_trace_scores] Page the scores attached to a trace. Required scopes: `traces:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `id` | string | Yes | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## list\_sessions [#list_sessions] Filter and page sessions; use list\_traces with sessionId to page their traces. Required scopes: `traces:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `filter` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## get\_session [#get_session] Read a session and its initial trace page. Required scopes: `traces:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `attributes` | object | Yes | | `createdAt` | string | Yes | | `id` | string | Yes | | `name` | null or string | Yes | | `nextCursor` | null or string | No | | `traceCount` | number | Yes | | `traces` | array of object | Yes | | `updatedAt` | string | Yes | ## list\_scorer\_versions [#list_scorer_versions] Page immutable scorer version metadata. Use get\_scorer\_version to retrieve executable configuration. Required scopes: `scorers:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `id` | string | Yes | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## get\_scorer\_version [#get_scorer_version] Read an immutable executable version belonging to a scorer. Required scopes: `scorers:read`. **Request fields** | Field | Type | Required | | ----------- | ------ | -------- | | `id` | string | Yes | | `versionId` | string | Yes | **Result** | Field | Type | Required | | ------------- | ------------------------ | -------- | | `code` | string | Yes | | `config` | JSON | No | | `createdAt` | string | Yes | | `evaluatorId` | string | Yes | | `id` | string | Yes | | `language` | "javascript" or "python" | Yes | | `version` | number | Yes | ## check\_scorer\_runtime [#check_scorer_runtime] Read-only configuration diagnostics for selected pinned scorers. Does not call providers or launch sandboxes, and does not establish connectivity or successful execution. Use probe\_scorer\_runtime explicitly for a bounded, potentially billable representative execution. Required scopes: `scorers:read`. **Request fields** | Field | Type | Required | | --------------------- | --------------- | -------- | | `scorerIds` | array of string | Yes | | `evaluatorVersionIds` | object | No | **Result** | Field | Type | Required | | ----------------------- | --------------- | -------- | | `checks` | array of object | Yes | | `incurredProviderUsage` | boolean | Yes | | `mode` | string | Yes | ## probe\_scorer\_runtime [#probe_scorer_runtime] Explicitly execute up to 3 pinned scorers on one representative trace/span or dataset case using their actual provider/model/sandbox. May incur usage and retain execution spans. Reports configuration, connectivity, and execution separately; creates no saved evaluation. Use start\_eval\_run for multi-case calibration. Required scopes: `scorers:read`, `scorers:write`, `traces:read`, `datasets:read`. **Request fields** | Field | Type | Required | | --------------------- | --------------- | -------- | | `scorerIds` | array of string | Yes | | `evaluatorVersionIds` | object | No | | `traceId` | string | Yes | | `spanId` | string | No | | `datasetId` | string | No | | `datasetItemId` | string | No | | `datasetVersionId` | string | No | **Result** | Field | Type | Required | | ---------------------- | --------------- | -------- | | `checks` | array of object | Yes | | `evaluationRunCreated` | boolean | Yes | | `mode` | string | Yes | ## test\_scorer [#test_scorer] Preview a saved, pinned, or inline scorer on trace or selected span evidence. Does not create a saved evaluation run or score result; a scorer execution span is retained. Use start\_eval\_run for a named persistent evaluation. LLM scorers may incur model cost. Optional dataset context requires both datasetId and datasetItemId. Required scopes: `scorers:read`, `scorers:write`, `traces:read`, `datasets:read`. **Request fields** | Field | Type | Required | | ------------------ | ------ | -------- | | `traceId` | string | Yes | | `spanId` | string | No | | `scorerId` | string | No | | `versionId` | string | No | | `scorer` | object | No | | `datasetId` | string | No | | `datasetItemId` | string | No | | `datasetVersionId` | string | No | **Result** | Field | Type | Required | | ----------------------- | ------- | -------- | | `evaluationRunCreated` | boolean | Yes | | `executionSpanRetained` | boolean | Yes | | `mode` | string | Yes | | `persisted` | boolean | Yes | | `result` | object | Yes | | `scorerId` | string | Yes | | `traceId` | string | Yes | | `versionId` | string | Yes | ## list\_eval\_runs [#list_eval_runs] Filter and page evaluation run summaries, including saved groups (type/name/version) and groupsResolvedAt. Filter membership with workflow = 'Name' or agent = 'Name'; multiple memberships are supported. Null groupsResolvedAt means historical attribution has not been resolved; an empty groups array with a timestamp means unassigned. Required scopes: `evals:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `filter` | string | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## list\_eval\_run\_groups [#list_eval_run_groups] Group all matching evaluation runs by saved workflow or agent membership, with exact run counts and cursor-paged groups. A mixed run can belong to several groups. Each returned filter can be passed to list\_eval\_runs to page that group's runs. Unresolved historical runs are separate from resolved, unassigned runs. Required scopes: `evals:read`. **Request fields** | Field | Type | Required | | -------------- | --------------------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `filter` | string | No | | `groupBy` | "workflow" or "agent" | Yes | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## cancel\_eval\_run [#cancel_eval_run] Stop scheduling work and fence late result writes. Already dispatched application calls may finish; retained outputs and judgments are preserved. Cancelled runs cannot be recovered. Required scopes: `evals:read`, `evals:write`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | --------------------- | ---------------------------------------------------------------- | -------- | | `completedAt` | null or string | Yes | | `createdAt` | string | Yes | | `datasetId` | null or string | Yes | | `evaluatorIds` | array of string | Yes | | `evaluatorVersionIds` | JSON | No | | `execution` | JSON | No | | `groups` | array of object | No | | `groupsResolvedAt` | null or string | No | | `id` | string | Yes | | `metadata` | object | Yes | | `name` | null or string | Yes | | `nextCursor` | null or string | No | | `resultCount` | number | Yes | | `results` | array of object | Yes | | `rows` | array of object | No | | `score` | null or number | Yes | | `scorerProgress` | array of object | No | | `scores` | array of object | No | | `status` | "cancelled" or "completed" or "failed" or "partial" or "running" | Yes | | `targetCount` | number | Yes | | `traceCount` | number | Yes | ## recover\_eval\_run [#recover_eval_run] Resume unfinished cases after a worker lease expires, or retry scoring errors on retained evidence. Reuses completed judgments and durable app/bridge output. Never re-dispatches an uncertain app call. Live workers and cancelled runs reject recovery. Required scopes: `evals:read`, `evals:write`, `traces:read`, `datasets:read`, `scorers:read`, `apps:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | --------------------- | ---------------------------------------------------------------- | -------- | | `completedAt` | null or string | Yes | | `createdAt` | string | Yes | | `datasetId` | null or string | Yes | | `evaluatorIds` | array of string | Yes | | `evaluatorVersionIds` | JSON | No | | `execution` | JSON | No | | `groups` | array of object | No | | `groupsResolvedAt` | null or string | No | | `id` | string | Yes | | `metadata` | object | Yes | | `name` | null or string | Yes | | `nextCursor` | null or string | No | | `resultCount` | number | Yes | | `results` | array of object | Yes | | `rows` | array of object | No | | `score` | null or number | Yes | | `scorerProgress` | array of object | No | | `scores` | array of object | No | | `status` | "cancelled" or "completed" or "failed" or "partial" or "running" | Yes | | `targetCount` | number | Yes | | `traceCount` | number | Yes | ## get\_eval\_run [#get_eval_run] Read a run's status, worker/stage progress, per-case execution errors, resolved evaluatorVersionIds and one lightweight result page with exact frozen inputs/outputs. Frozen and scorer spans are omitted by default; use get\_eval\_target with the run id and a row id for full case evidence, or opt in with includeEvidence. Batches shrink automatically to fit 8 MiB; continue with nextCursor, even if fewer rows than requested are returned. Required scopes: `evals:read`. **Request fields** | Field | Type | Required | | ----------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `id` | string | Yes | | `includeEvidence` | boolean | No | **Result** | Field | Type | Required | | --------------------- | ---------------------------------------------------------------- | -------- | | `completedAt` | null or string | Yes | | `createdAt` | string | Yes | | `datasetId` | null or string | Yes | | `evaluatorIds` | array of string | Yes | | `evaluatorVersionIds` | JSON | No | | `execution` | JSON | No | | `groups` | array of object | No | | `groupsResolvedAt` | null or string | No | | `id` | string | Yes | | `metadata` | object | Yes | | `name` | null or string | Yes | | `nextCursor` | null or string | No | | `resultCount` | number | Yes | | `results` | array of object | Yes | | `rows` | array of object | No | | `score` | null or number | Yes | | `scorerProgress` | array of object | No | | `scores` | array of object | No | | `status` | "cancelled" or "completed" or "failed" or "partial" or "running" | Yes | | `targetCount` | number | Yes | | `traceCount` | number | Yes | ## get\_eval\_target [#get_eval_target] Read one evaluation case's complete frozen evidence, scorer execution spans and results, capped at 8 MiB. Pass the run id and targetId from get\_eval\_run rows\[].id or a comparison row's id. Does not substitute the current live trace for saved evidence. Required scopes: `evals:read`. **Request fields** | Field | Type | Required | | ---------- | ------ | -------- | | `id` | string | Yes | | `targetId` | string | Yes | **Result** | Field | Type | Required | | ---------------- | ------------------------------------------------------------------------------------------------------------------------------------ | -------- | | `datasetCaseId` | null or string | No | | `datasetItemId` | null or string | Yes | | `executionError` | null or string | No | | `expectedOutput` | null or string or number or false or true or object or array of null or string or number or false or true or object or array of JSON | Yes | | `id` | string | Yes | | `results` | array of object | Yes | | `scorerStatuses` | JSON | No | | `scoringTrace` | JSON | No | | `sourceSpanId` | null or string | No | | `stage` | string | No | | `trace` | object | Yes | ## start\_eval\_run [#start_eval_run] Create a named, persistent asynchronous evaluation and return its URL promptly. Execution completion, quality failures, and infrastructure errors are distinct. Choose trace IDs, a dataset (optionally datasetVersionId), or sourceRunId to re-score frozen evidence without calling the app again. Connected dataset runs accept promptOverrides keyed by managed prompt slug with optional version and model; all published prompt defaults are frozen at run creation, including lazy discovery. These are separate from inputs and expected outputs, forbidden for re-scoring, and part of requestKey identity. Connected dataset runs accept inputOverrides: shallow merge over each object case input, override keys win, nested values replace whole values and null is literal. The effective app input and overrides are frozen; dataset inputs and expected outputs remain unchanged. Overrides are forbidden for re-scoring and included in requestKey identity. Use parentRunId to execute the app again on a prior run's frozen cases/references, inheriting its settings and resolving latest prompts/scorers by default; optional overrides change those settings. useRecordedVersions=true holds recorded prompt/scorer revisions for reproduction or comparison. sourceRunId only re-scores saved outputs; its judges also default to latest. Each run records resolved versions, parentRunId and configurationChanges. Pin individual scorers with evaluatorVersionIds. Supply a stable requestKey: retries with the same key and inputs return the same run. Connected runs and LLM scorers may incur cost. Poll get\_eval\_run and use gate\_eval\_run when terminal. Required scopes: `evals:write`, `evals:read`, `traces:read`, `datasets:read`, `scorers:read`. **Request fields** | Field | Type | Required | | --------------------- | ----------------------- | -------- | | `parentRunId` | string | No | | `useRecordedVersions` | boolean | No | | `promptOverrides` | object | No | | `inputOverrides` | object | No | | `datasetVersionId` | string | No | | `evaluatorVersionIds` | object | No | | `sourceRunId` | string | No | | `background` | boolean | No | | `mode` | "connected" or "traces" | No | | `appId` | string | No | | `concurrency` | integer | No | | `metadata` | object | No | | `datasetId` | string | No | | `datasetItemIds` | array of string | No | | `evaluatorIds` | array of string | No | | `name` | string | No | | `traceIds` | array of string | No | | `requestKey` | string | Yes | **Result** | Field | Type | Required | | --------------------- | ---------------------------------------------------------------- | -------- | | `completedAt` | null or string | Yes | | `createdAt` | string | Yes | | `datasetId` | null or string | Yes | | `evaluatorIds` | array of string | Yes | | `evaluatorVersionIds` | JSON | No | | `execution` | JSON | No | | `groups` | array of object | No | | `groupsResolvedAt` | null or string | No | | `id` | string | Yes | | `metadata` | object | Yes | | `name` | null or string | Yes | | `nextCursor` | null or string | No | | `resultCount` | number | Yes | | `results` | array of object | Yes | | `rows` | array of object | No | | `score` | null or number | Yes | | `scorerProgress` | array of object | No | | `scores` | array of object | No | | `status` | "cancelled" or "completed" or "failed" or "partial" or "running" | Yes | | `targetCount` | number | Yes | | `traceCount` | number | Yes | | `url` | string | Yes | ## compare\_eval\_runs [#compare_eval_runs] Compare runs and their resolved configurationChanges, separating extractor changes from judge-version changes. Changed judges remain viewable but cannot pass a baseline quality gate. Use exact frozen inputs/outputs, matching dataset item, trace, then unique input. Full spans are omitted by default; use get\_eval\_target to inspect a case or opt in with includeEvidence. Batches shrink automatically to fit 8 MiB; always continue with nextOffset instead of assuming a fixed page size. Required scopes: `evals:read`. **Request fields** | Field | Type | Required | | ----------------- | ------- | -------- | | `leftId` | string | Yes | | `rightId` | string | Yes | | `offset` | integer | No | | `includeEvidence` | boolean | No | **Result** | Field | Type | Required | | ---------------------- | --------------- | -------- | | `configurationChanges` | array of object | Yes | | `left` | object | Yes | | `limit` | number | Yes | | `nextOffset` | null or number | Yes | | `offset` | number | Yes | | `pairs` | array of object | Yes | | `right` | object | Yes | | `total` | number | Yes | ## gate\_eval\_run [#gate_eval_run] Evaluate CI thresholds across ALL results. Fails closed on nonterminal, empty, missing or unscored results. Optional baseline checks mean-score regression. Returns passed plus reasons; CLI exits 2 on failure. Required scopes: `evals:read`. **Request fields** | Field | Type | Required | | --------------- | ------- | -------- | | `id` | string | Yes | | `minScore` | number | No | | `minPassRate` | number | No | | `maxErrors` | integer | No | | `allowUnscored` | boolean | No | | `baselineId` | string | No | | `maxRegression` | number | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `baseline` | null or object | Yes | | `passed` | boolean | Yes | | `reasons` | array of string | Yes | | `run` | object | Yes | | `thresholds` | object | Yes | ## promote\_spans [#promote_spans] Preview or atomically promote 1–100 exact production spans into dataset cases. Defaults to preview=true. Captures only the selected subtree with provenance; observed output is separate and expectedOutput defaults to null. mappedInput is explicit app input, never inferred from prompt text. Save with preview=false and the preview's expectedEvidenceHash. No new production traces are created. Required scopes: `datasets:read`, `datasets:write`, `traces:read`. **Request fields** | Field | Type | Required | | ---------------------- | --------------- | -------- | | `datasetId` | string | Yes | | `preview` | boolean | No | | `expectedHash` | string | No | | `expectedEvidenceHash` | string | No | | `spans` | array of object | Yes | **Result** | Field | Type | Required | | -------------- | --------------- | -------- | | `cases` | array of object | Yes | | `created` | array of string | Yes | | `datasetId` | string | Yes | | `evidenceHash` | string | Yes | | `preview` | boolean | Yes | ## bulk\_dataset\_items [#bulk_dataset_items] Atomically create, update and delete at most 100 dataset items. Any failure rolls back the batch. expectedHash optionally guards the content hash returned by a snapshot or prior bulk edit. Required scopes: `datasets:write`. **Request fields** | Field | Type | Required | | -------------- | --------------- | -------- | | `datasetId` | string | Yes | | `expectedHash` | string | No | | `create` | array of object | No | | `update` | array of object | No | | `delete` | array of string | No | **Result** | Field | Type | Required | | ------------- | --------------- | -------- | | `contentHash` | string | Yes | | `created` | array of string | Yes | | `datasetId` | string | Yes | | `deleted` | array of string | Yes | | `updated` | array of string | Yes | ## create\_dataset\_snapshot [#create_dataset_snapshot] Freeze dataset inputs, expected outputs, metadata and captured span evidence as immutable per-case content. Aggregate captured evidence can exceed 8 MiB; item reads remain bounded to 8 MiB per page. Legacy trace references resolve evidence at evaluation start. Identical content returns the existing snapshot. Pass its id as datasetVersionId to an evaluation. Required scopes: `datasets:read`, `datasets:write`. **Request fields** | Field | Type | Required | | ----------- | ------ | -------- | | `datasetId` | string | Yes | | `label` | string | No | **Result** | Field | Type | Required | | ------------- | -------------- | -------- | | `contentHash` | string | Yes | | `createdAt` | string | Yes | | `datasetId` | string | Yes | | `id` | string | Yes | | `itemCount` | number | Yes | | `label` | null or string | Yes | ## list\_dataset\_snapshots [#list_dataset_snapshots] Page immutable dataset snapshot metadata. Required scopes: `datasets:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `datasetId` | string | Yes | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `items` | array of object | Yes | | `nextCursor` | null or string | Yes | | `total` | number | No | ## get\_dataset\_snapshot [#get_dataset_snapshot] Read a frozen dataset and one immutable item page. Pages shrink to fit 8 MiB; continue with nextCursor even when fewer rows than requested are returned. Required scopes: `datasets:read`. **Request fields** | Field | Type | Required | | -------------- | ------- | -------- | | `includeTotal` | boolean | No | | `limit` | integer | No | | `cursor` | string | No | | `datasetId` | string | Yes | | `versionId` | string | Yes | **Result** | Field | Type | Required | | -------------- | --------------- | -------- | | `createdAt` | string | Yes | | `description` | null or string | Yes | | `fieldSchemas` | JSON | No | | `id` | string | Yes | | `itemCount` | number | Yes | | `items` | array of object | Yes | | `metadata` | JSON | No | | `name` | string | Yes | | `nextCursor` | null or string | Yes | | `revision` | number | No | | `updatedAt` | string | Yes | | `versionId` | string | Yes | ## preview\_dashboard [#preview_dashboard] Execute stored dashboard widget queries with their saved date windows. Returns configuration and ordered query results. Each batch of at most 40 expanded queries shares a snapshot; larger previews use multiple snapshots, not one dashboard-wide snapshot. On READ\_BUSY, honor details.retryAfterSeconds and retry this read with bounded backoff. Required scopes: `dashboards:read`, `metrics:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `id` | string | Yes | **Result** | Field | Type | Required | | ------------- | --------------- | -------- | | `comparisons` | array of object | Yes | | `dashboard` | object | Yes | | `results` | array of object | Yes | ## get\_saved\_view\_data [#get_saved_view_data] Execute a saved selector view and return a bounded row page. Eval-result views can be restricted to runId. Required scopes: `views:read`, `traces:read`, `evals:read`. **Request fields** | Field | Type | Required | | -------- | ------- | -------- | | `id` | string | Yes | | `runId` | string | No | | `offset` | integer | No | | `limit` | integer | No | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `columns` | array of object | Yes | | `limit` | number | Yes | | `nextOffset` | null or number | Yes | | `offset` | number | Yes | | `rows` | array of object | Yes | | `total` | number | Yes | | `view` | object | Yes | ## list\_report\_templates [#list_report_templates] List query-backed MDX report starters. get\_report\_template returns an editable MDX document with named sources. Required scopes: `dashboards:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** — array; each item has these fields | Field | Type | Required | | ------------- | ------ | -------- | | `description` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | ## get\_report\_template [#get_report_template] Get an MDX starter document with concrete named queries and absolute dates. asOf sets the query window end, not a historical database snapshot. Review sources and compose MDX before creating a report. Required scopes: `dashboards:read`. **Request fields** | Field | Type | Required | | ------ | ------ | -------- | | `id` | string | Yes | | `asOf` | string | No | **Result** | Field | Type | Required | | ------------- | ------ | -------- | | `description` | string | Yes | | `document` | object | Yes | | `id` | string | Yes | | `name` | string | Yes | ## list\_reports [#list_reports] List saved frozen reports in the current project, newest first. Use get\_report with the project-local report number to read its config and captured data. Required scopes: `dashboards:read`, `metrics:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** — array; each item has these fields | Field | Type | Required | | ------------- | ---------------------- | -------- | | `author` | null or object | No | | `createdAt` | string | Yes | | `description` | string | Yes | | `frozenAt` | string | Yes | | `id` | string | Yes | | `name` | string | Yes | | `number` | number | Yes | | `publishedAt` | null or string | No | | `revision` | number | No | | `status` | "draft" or "published" | No | | `templateId` | string | Yes | | `updatedAt` | null or string | No | | `widgetCount` | number | Yes | ## get\_report [#get_report] Read a saved report by its project-local number, including its text, widget config and frozen query results. This does not query live metric data. Required scopes: `dashboards:read`, `metrics:read`. **Request fields** | Field | Type | Required | | -------- | ------- | -------- | | `number` | integer | Yes | **Result** | Field | Type | Required | | ---------------------- | ---------------------------------------------------------------- | -------- | | `author` | null or object | No | | `config` | object | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `document` | JSON | No | | `draftInput` | JSON | No | | `frozenAt` | string | Yes | | `id` | string | Yes | | `layout` | "comparison" or "document" or "canvas" or "brief" or "scorecard" | No | | `mdx` | JSON | No | | `name` | string | Yes | | `number` | number | Yes | | `presentation` | null or object | No | | `presentationRevision` | number | No | | `publicPath` | null or string | No | | `publicUrl` | null or string | No | | `publishedAt` | null or string | No | | `revision` | number | No | | `snapshot` | object | Yes | | `status` | "draft" or "published" | No | | `templateId` | string | Yes | | `updatedAt` | null or string | No | | `widgetCount` | number | Yes | ## create\_report [#create_report] Create a private MDX report draft from name, mdx, named sources and optional evidence bindings. Start with get\_report\_authoring\_guide for the workflow, paired-file recipe, exact component props and supported Tailwind utilities. Component source props reference named queries. No JavaScript execution, imports or arbitrary CSS. Use literal JSON props. All queries capture complete results together (40 queries, 5000 rows/query, 8 MiB); sources reused by components are deduplicated. Validate with validate\_report first. Creation requires a stable UUID creationKey; identical retries return the original. update\_report preserves captured data unless refresh is explicit. publish\_report locks the reviewed revision; sharing remains a separate action. Required scopes: `dashboards:write`, `metrics:read`. **Request fields** | Field | Type | Required | | ------------- | ------ | -------- | | `name` | string | Yes | | `description` | string | No | | `mdx` | string | Yes | | `sources` | object | No | | `bindings` | object | No | | `creationKey` | string | Yes | **Result** | Field | Type | Required | | ---------------------- | ---------------------------------------------------------------- | -------- | | `author` | null or object | No | | `config` | object | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `document` | JSON | No | | `draftInput` | JSON | No | | `frozenAt` | string | Yes | | `id` | string | Yes | | `layout` | "comparison" or "document" or "canvas" or "brief" or "scorecard" | No | | `mdx` | JSON | No | | `name` | string | Yes | | `number` | number | Yes | | `presentation` | null or object | No | | `presentationRevision` | number | No | | `publicPath` | null or string | No | | `publicUrl` | null or string | No | | `publishedAt` | null or string | No | | `revision` | number | No | | `snapshot` | object | Yes | | `status` | "draft" or "published" | No | | `templateId` | string | Yes | | `updatedAt` | null or string | No | | `widgetCount` | number | Yes | ## get\_report\_recipe [#get_report_recipe] Get the MDX evaluation-story recipe: editorial guidance, report.mdx and report.data.json starter files, and the agent workflow. Use get\_report\_components for typed props. The recipe is optional guidance, not a fixed layout. Required scopes: `dashboards:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** | Field | Type | Required | | ------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | -------- | | `description` | "A concise opening, evidence-bound metrics, progression, candidate comparison, expandable supporting evidence and a closing scorecard. Adapt the composition to the question." | Yes | | `files` | object | Yes | | `format` | "mdx" | Yes | | `id` | "evaluation-story" | Yes | | `name` | "Evaluation story" | Yes | | `workflow` | array of JSON | Yes | ## get\_report\_authoring\_guide [#get_report_authoring_guide] Get the complete MDX report authoring contract: document shape, source and binding model, component examples, limits, validation scope and the ordered workflow from discovery through private draft review and optional publication. Use this before writing a report; it does not query data or save a report. Required scopes: `dashboards:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** | Field | Type | Required | | ------------------- | ----------------------------------------------------------------------------------------------------------------- | -------- | | `cli` | array of JSON | Yes | | `componentCatalog` | object | Yes | | `componentExamples` | array of JSON | Yes | | `contract` | object | Yes | | `format` | "mdx" | Yes | | `id` | "mdx-report-authoring" | Yes | | `limits` | object | Yes | | `name` | "MDX report authoring" | Yes | | `nextSteps` | object | Yes | | `preview` | object | Yes | | `purpose` | "Compose a private, evidence-bound report draft from Markdown, registered components and named semantic queries." | Yes | | `recipe` | object | Yes | | `validationScope` | object | Yes | | `version` | 1 | Yes | | `workflow` | array of JSON | Yes | ## get\_report\_components [#get_report_components] Discover MDX component names, exact prop schemas, child support and supported static Tailwind classes. Required scopes: `dashboards:read`. **Request fields** | Field | Type | Required | | ----- | ---- | -------- | **Result** | Field | Type | Required | | ------------ | --------------- | -------- | | `classes` | array of string | Yes | | `components` | array of object | Yes | | `htmlTags` | array of string | Yes | | `syntax` | string | Yes | | `version` | number | Yes | ## validate\_report [#validate_report] Validate MDX, component props, sources, evidence and data shapes without saving or consuming a report number. With number and revision, validates against that report's frozen evidence; refresh:true explicitly queries fresh data. Without number, performs a fresh capture for validation. Returns structural, data and server-render checks; browser visual and responsive review remain separate. Does not publish or share. Required scopes: `dashboards:read`, `metrics:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `document` | object | Yes | | `number` | integer | No | | `revision` | integer | No | | `refresh` | boolean | No | **Result** Type: **object**. ## update\_report [#update_report] Edit a private MDX draft using number, revision and the complete document \{name, description, mdx, sources, bindings}. Source and layout edits reuse frozen evidence. Changing query definitions or adding a visualization requiring extra aggregates requires refresh:true; review the refreshed data before publishing. Published reports reject edits. Required scopes: `dashboards:write`, `metrics:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `number` | integer | Yes | | `revision` | integer | Yes | | `document` | object | Yes | | `refresh` | boolean | No | **Result** | Field | Type | Required | | ---------------------- | ---------------------------------------------------------------- | -------- | | `author` | null or object | No | | `config` | object | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `document` | JSON | No | | `draftInput` | JSON | No | | `frozenAt` | string | Yes | | `id` | string | Yes | | `layout` | "comparison" or "document" or "canvas" or "brief" or "scorecard" | No | | `mdx` | JSON | No | | `name` | string | Yes | | `number` | number | Yes | | `presentation` | null or object | No | | `presentationRevision` | number | No | | `publicPath` | null or string | No | | `publicUrl` | null or string | No | | `publishedAt` | null or string | No | | `revision` | number | No | | `snapshot` | object | Yes | | `status` | "draft" or "published" | No | | `templateId` | string | Yes | | `updatedAt` | null or string | No | | `widgetCount` | number | Yes | ## publish\_report [#publish_report] Publish the reviewed draft at its current revision. Locks the composition and captured evidence without querying live data. Publication alone stays private; use set\_report\_sharing only when public access is intended. Required scopes: `dashboards:write`, `metrics:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `number` | integer | Yes | | `revision` | integer | Yes | **Result** | Field | Type | Required | | ---------------------- | ---------------------------------------------------------------- | -------- | | `author` | null or object | No | | `config` | object | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `document` | JSON | No | | `draftInput` | JSON | No | | `frozenAt` | string | Yes | | `id` | string | Yes | | `layout` | "comparison" or "document" or "canvas" or "brief" or "scorecard" | No | | `mdx` | JSON | No | | `name` | string | Yes | | `number` | number | Yes | | `presentation` | null or object | No | | `presentationRevision` | number | No | | `publicPath` | null or string | No | | `publicUrl` | null or string | No | | `publishedAt` | null or string | No | | `revision` | number | No | | `snapshot` | object | Yes | | `status` | "draft" or "published" | No | | `templateId` | string | Yes | | `updatedAt` | null or string | No | | `widgetCount` | number | Yes | ## set\_report\_sharing [#set_report_sharing] Explicitly enable or revoke an anonymous public link for a published report. Anyone with the returned publicPath can read all captured report rows and expanded evidence without signing in. Never enable without authorization to share that report publicly. Revocation invalidates the link; re-enabling generates a new one. Required scopes: `dashboards:write`, `metrics:read`. **Request fields** | Field | Type | Required | | ---------- | ------- | -------- | | `number` | integer | Yes | | `revision` | integer | Yes | | `enabled` | boolean | Yes | **Result** | Field | Type | Required | | ---------------------- | ---------------------------------------------------------------- | -------- | | `author` | null or object | No | | `config` | object | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `document` | JSON | No | | `draftInput` | JSON | No | | `frozenAt` | string | Yes | | `id` | string | Yes | | `layout` | "comparison" or "document" or "canvas" or "brief" or "scorecard" | No | | `mdx` | JSON | No | | `name` | string | Yes | | `number` | number | Yes | | `presentation` | null or object | No | | `presentationRevision` | number | No | | `publicPath` | null or string | No | | `publicUrl` | null or string | No | | `publishedAt` | null or string | No | | `revision` | number | No | | `snapshot` | object | Yes | | `status` | "draft" or "published" | No | | `templateId` | string | Yes | | `updatedAt` | null or string | No | | `widgetCount` | number | Yes | ## clone\_report [#clone_report] Create a new private draft copy of a report, preserving the exact captured evidence without rerunning queries. Uses a UUID creationKey for idempotent retries; the original and any public link stay unchanged. Required scopes: `dashboards:write`, `metrics:read`. **Request fields** | Field | Type | Required | | ------------- | ------- | -------- | | `number` | integer | Yes | | `creationKey` | string | Yes | **Result** | Field | Type | Required | | ---------------------- | ---------------------------------------------------------------- | -------- | | `author` | null or object | No | | `config` | object | Yes | | `createdAt` | string | Yes | | `description` | string | Yes | | `document` | JSON | No | | `draftInput` | JSON | No | | `frozenAt` | string | Yes | | `id` | string | Yes | | `layout` | "comparison" or "document" or "canvas" or "brief" or "scorecard" | No | | `mdx` | JSON | No | | `name` | string | Yes | | `number` | number | Yes | | `presentation` | null or object | No | | `presentationRevision` | number | No | | `publicPath` | null or string | No | | `publicUrl` | null or string | No | | `publishedAt` | null or string | No | | `revision` | number | No | | `snapshot` | object | Yes | | `status` | "draft" or "published" | No | | `templateId` | string | Yes | | `updatedAt` | null or string | No | | `widgetCount` | number | Yes | ## resolve\_trace [#resolve_trace] Resolve an exact ID or supported exact name/slug in the current project and return its canonical URL. Reports also resolve by project-local number. Ambiguous names fail. Some objects have collection links because the UI has no detail route. Required scopes: `traces:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `key` | string | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `kind` | "trace" or "session" or "dataset" or "scorer" or "eval" or "dashboard" or "report" or "view" | Yes | | `linkKind` | string | Yes | | `path` | string | Yes | | `projectId` | string | Yes | | `url` | null or string | Yes | ## resolve\_session [#resolve_session] Resolve an exact ID or supported exact name/slug in the current project and return its canonical URL. Reports also resolve by project-local number. Ambiguous names fail. Some objects have collection links because the UI has no detail route. Required scopes: `traces:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `key` | string | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `kind` | "trace" or "session" or "dataset" or "scorer" or "eval" or "dashboard" or "report" or "view" | Yes | | `linkKind` | string | Yes | | `path` | string | Yes | | `projectId` | string | Yes | | `url` | null or string | Yes | ## resolve\_dataset [#resolve_dataset] Resolve an exact ID or supported exact name/slug in the current project and return its canonical URL. Reports also resolve by project-local number. Ambiguous names fail. Some objects have collection links because the UI has no detail route. Required scopes: `datasets:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `key` | string | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `kind` | "trace" or "session" or "dataset" or "scorer" or "eval" or "dashboard" or "report" or "view" | Yes | | `linkKind` | string | Yes | | `path` | string | Yes | | `projectId` | string | Yes | | `url` | null or string | Yes | ## resolve\_scorer [#resolve_scorer] Resolve an exact ID or supported exact name/slug in the current project and return its canonical URL. Reports also resolve by project-local number. Ambiguous names fail. Some objects have collection links because the UI has no detail route. Required scopes: `scorers:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `key` | string | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `kind` | "trace" or "session" or "dataset" or "scorer" or "eval" or "dashboard" or "report" or "view" | Yes | | `linkKind` | string | Yes | | `path` | string | Yes | | `projectId` | string | Yes | | `url` | null or string | Yes | ## resolve\_eval [#resolve_eval] Resolve an exact ID or supported exact name/slug in the current project and return its canonical URL. Reports also resolve by project-local number. Ambiguous names fail. Some objects have collection links because the UI has no detail route. Required scopes: `evals:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `key` | string | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `kind` | "trace" or "session" or "dataset" or "scorer" or "eval" or "dashboard" or "report" or "view" | Yes | | `linkKind` | string | Yes | | `path` | string | Yes | | `projectId` | string | Yes | | `url` | null or string | Yes | ## resolve\_dashboard [#resolve_dashboard] Resolve an exact ID or supported exact name/slug in the current project and return its canonical URL. Reports also resolve by project-local number. Ambiguous names fail. Some objects have collection links because the UI has no detail route. Required scopes: `dashboards:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `key` | string | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `kind` | "trace" or "session" or "dataset" or "scorer" or "eval" or "dashboard" or "report" or "view" | Yes | | `linkKind` | string | Yes | | `path` | string | Yes | | `projectId` | string | Yes | | `url` | null or string | Yes | ## resolve\_view [#resolve_view] Resolve an exact ID or supported exact name/slug in the current project and return its canonical URL. Reports also resolve by project-local number. Ambiguous names fail. Some objects have collection links because the UI has no detail route. Required scopes: `views:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `key` | string | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `kind` | "trace" or "session" or "dataset" or "scorer" or "eval" or "dashboard" or "report" or "view" | Yes | | `linkKind` | string | Yes | | `path` | string | Yes | | `projectId` | string | Yes | | `url` | null or string | Yes | ## resolve\_report [#resolve_report] Resolve an exact ID or supported exact name/slug in the current project and return its canonical URL. Reports also resolve by project-local number. Ambiguous names fail. Some objects have collection links because the UI has no detail route. Required scopes: `dashboards:read`, `metrics:read`. **Request fields** | Field | Type | Required | | ----- | ------ | -------- | | `key` | string | Yes | **Result** | Field | Type | Required | | ----------- | -------------------------------------------------------------------------------------------- | -------- | | `id` | string | Yes | | `kind` | "trace" or "session" or "dataset" or "scorer" or "eval" or "dashboard" or "report" or "view" | Yes | | `linkKind` | string | Yes | | `path` | string | Yes | | `projectId` | string | Yes | | `url` | null or string | Yes | ## describe\_agent\_operations [#describe_agent_operations] Discover agent operation names, required permissions and JSON input schemas shared by MCP and CLI. Required scopes: no additional operation scopes (authentication and project access still required). **Request fields** | Field | Type | Required | | ------ | ------ | -------- | | `name` | string | No | **Result** — array; each item has these fields | Field | Type | Required | | ------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------- | | `description` | string | Yes | | `inputSchema` | object | Yes | | `name` | string | Yes | | `scopes` | array of "apps:read" or "apps:write" or "reviews:read" or "reviews:write" or "metrics:read" or "scorers:read" or "scorers:write" or "views:read" or "views:write" or "datasets:read" or "datasets:write" or "dashboards:read" or "dashboards:write" or "traces:read" or "evals:read" or "evals:write" or "prompts:read" or "prompts:write" or "traces:write" or "playgrounds:read" or "playgrounds:write" | Yes |