# Datool

> Trace AI applications, inspect their behavior, and evaluate their outputs.

These are the public product docs. Each link returns Markdown. Credentials and project access are required for product API operations, but not for these docs.

## Documentation

- [Documentation](https://www.trydatool.com/docs.md): Understand what your AI application did, find what went wrong, and measure improvements.
- [CI evaluation gates](https://www.trydatool.com/docs/evaluation/ci.md): Run a saved evaluation, wait for execution, and fail a build on explicit quality thresholds.
- [Datasets](https://www.trydatool.com/docs/evaluation/datasets.md): Turn representative inputs and known failures into repeatable evaluation cases.
- [Evaluation runs](https://www.trydatool.com/docs/evaluation/runs.md): Execute dataset cases against an app and compare the resulting evidence and scores.
- [Scorers](https://www.trydatool.com/docs/evaluation/scorers.md): Define how to judge an output, test the rule, and save a version for evaluations.
- [Core concepts](https://www.trydatool.com/docs/get-started/concepts.md): How project data, recorded behavior, and evaluations fit together.
- [Run your first evaluation](https://www.trydatool.com/docs/get-started/first-evaluation.md): Connect a small app, catch a failing case, fix it, and compare two saved runs without a model API key.
- [Use your first managed prompt](https://www.trydatool.com/docs/get-started/first-prompt.md): Publish a prompt and use its latest published version from your application.
- [Send your first trace](https://www.trydatool.com/docs/get-started/first-trace.md): Trace an AI SDK call, a TypeScript or Python function, or a LangGraph application.
- [Alerts](https://www.trydatool.com/docs/guides/alerts.md): Configure request-failure alerts, count repeated errors, and diagnose webhook delivery.
- [Dashboards](https://www.trydatool.com/docs/guides/dashboards.md): Build a weekly health dashboard, choose the right data source, and interpret changes without mixing populations.
- [Playground](https://www.trydatool.com/docs/guides/playground.md): Connect an application, try inputs, and inspect its outputs and traces.
- [Managed prompts](https://www.trydatool.com/docs/guides/prompts.md): Publish message templates and fetch the latest published configuration from your application.
- [Model and sandbox providers](https://www.trydatool.com/docs/guides/providers.md): Configure the credentials and execution environments used by prompt previews and scorers.
- [Reviews and Human Scores](https://www.trydatool.com/docs/guides/reviews.md): Collect structured judgments on recorded traces and preserve the reasoning behind them.
- [REST API](https://www.trydatool.com/docs/reference/api.md): Call project operations with typed inputs and results, explicit pagination, and structured errors.
- [Authentication and project access](https://www.trydatool.com/docs/reference/authentication.md): Choose browser sessions, organization API keys, or user OAuth for the task.
- [CLI reference](https://www.trydatool.com/docs/reference/cli.md): Authenticate, inspect traces, connect applications, and run evaluations from a terminal.
- [Versions and compatibility](https://www.trydatool.com/docs/reference/compatibility.md): Choose installable packages and distinguish public releases from preview integrations.
- [Filters and saved views](https://www.trydatool.com/docs/reference/filters.md): Search traces, sessions, and evaluations with typed comparisons and reusable URL filters.
- [MCP reference](https://www.trydatool.com/docs/reference/mcp.md): Connect an AI assistant to a project through Datool's authenticated MCP server.
- [Operation reference](https://www.trydatool.com/docs/reference/operations.md): Browse request fields, response fields, and required scopes for every REST, CLI, and MCP operation.
- [Python SDK](https://www.trydatool.com/docs/reference/python-sdk.md): Trace Python functions, async applications, and streams with Datool's native Python SDK.
- [SDK reference](https://www.trydatool.com/docs/reference/sdk.md): Entry points, configuration, and delivery behavior of the Node.js SDK.
- [Troubleshooting](https://www.trydatool.com/docs/reference/troubleshooting.md): Diagnose missing traces, unavailable apps, prompt reads, evaluation failures, and empty dashboards.
- [Self-hosting](https://www.trydatool.com/docs/self-hosting.md): Run Datool with PostgreSQL, Redis, and a persistent ingestion worker.
- [AI SDK](https://www.trydatool.com/docs/tracing/ai-sdk.md): Record a complete AI SDK 7 model call with Datool's OpenTelemetry processor.
- [Hermes Agent](https://www.trydatool.com/docs/tracing/hermes.md): Automatically collect Hermes turns, model calls and tool results in Datool using a native plugin.
- [Instrumentation](https://www.trydatool.com/docs/tracing/instrumentation.md): Record manual operations or export existing OpenTelemetry spans to Datool.
- [Investigate traces](https://www.trydatool.com/docs/tracing/investigation.md): Follow an execution, inspect its evidence, and keep useful examples for evaluation.
- [LangGraph and LangChain](https://www.trydatool.com/docs/tracing/langgraph.md): Record a complete Python graph with Datool callbacks, including nested steps and their inputs and outputs.
- [OpenAI Chat Completions](https://www.trydatool.com/docs/tracing/openai.md): Trace a non-streaming Chat Completions request through the OpenAI client's fetch hook.

## API discovery

- [OpenAPI specification](https://www.trydatool.com/openapi.json): Agent operations, request schemas, required scopes, and errors.
- [OAuth authorization server](https://www.trydatool.com/.well-known/oauth-authorization-server): Authorization code with PKCE and refresh tokens.
- [MCP protected resource](https://www.trydatool.com/.well-known/oauth-protected-resource/api/mcp): MCP authentication and scopes.