> ## Documentation Index
> Fetch the complete documentation index at: https://chatobserver.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# How can an AI assistant read my saved chatobserver data through MCP?

> Discover saved queries, analysis views and recent runs through the authenticated chatobserver Data MCP endpoint.

The Data MCP endpoint gives an authenticated assistant read access to saved queries, analysis views, and recent runs visible to its service account. It does not create queries, execute new runs, or return run result rows through these tools.

## Connect with a scoped key

Use `POST /api/v2/metrics/mcp` with a service-account API key carrying `metrics:read`. The same workspace and connection permissions used by the Data API apply. Keep the key in the assistant application's secure connection settings.

Send JSON-RPC 2.0 requests for `initialize`, `tools/list`, or `tools/call`. Discover the available tool schemas with `tools/list` before constructing a call.

## Choose a read operation

| Tool                          | Purpose                                       |
| ----------------------------- | --------------------------------------------- |
| `metrics_list_queries`        | List visible saved queries.                   |
| `metrics_get_query`           | Get a visible saved query by ID.              |
| `metrics_list_runs`           | List recent runs.                             |
| `metrics_get_run`             | Get one recent run by ID.                     |
| `metrics_list_analysis_views` | List reusable analysis views.                 |
| `metrics_get_analysis_view`   | Get an analysis view and its version history. |

List operations return a bounded set of records. The query and run get tools search the visible recent collection; use the REST API for paginated workflows and operation results.

## Keep discovery separate from execution

Use these tools to understand existing definitions and results before taking an action elsewhere. To create a query or execute an operation, implement the corresponding [Data API](/docs/developers/data-api) command with its required scope and idempotency key. See [operations and webhooks](/docs/developers/data-operations) for completion handling.