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

# API and MCP

> Read Agent Picks measurements and manage model choices and schedules from your own tools.

Agent Picks is available through the [REST v2 API](/api-reference/overview) and [MCP server](/api-reference/mcp-server). Use these when your own agent or reporting tool needs the same measurements you read in the app.

Start with [API authentication](/api-reference/authentication), or connect an agent through the MCP guide. Access requires Agent Picks on the account and access to the project. Reads require `visibility.read`; model and schedule changes require `actions.write`. Schedule changes also require project-management permission.

## Endpoints and tools

REST paths below are relative to `https://devtune.ai/api/v2`. Follow each reference link for parameters and response fields.

| REST v2 reference | MCP tool |
| - | - |
| [`GET /projects/{projectId}/agent-picks/summary`](/api-reference/get-agent-picks-summary) | `devtune_get_agent_picks_summary` |
| [`GET /projects/{projectId}/agent-picks/runs`](/api-reference/list-agent-picks-runs) | `devtune_list_agent_picks_runs` |
| [`GET /projects/{projectId}/agent-picks/tests/{testId}/runs/{runId}`](/api-reference/get-agent-picks-run-evidence) | `devtune_get_agent_picks_run` |
| [`GET /projects/{projectId}/agent-picks/starter-test`](/api-reference/get-agent-picks-starter-test) | `devtune_get_agent_picks_starter_test` |
| [`GET /projects/{projectId}/agent-picks/test`](/api-reference/get-paid-agent-picks-test) | `devtune_get_agent_picks_test` |
| [`GET /projects/{projectId}/agent-picks/history`](/api-reference/get-agent-picks-test-history) | `devtune_get_agent_picks_history` |
| [`GET /projects/{projectId}/agent-picks/actions`](/api-reference/get-agent-picks-actions-and-re-test-results) | `devtune_get_agent_picks_actions` |
| [`PATCH /projects/{projectId}/agent-picks/settings`](/api-reference/update-agent-picks-model-settings) | `devtune_update_agent_picks_settings` |
| [`GET /projects/{projectId}/agent-picks/schedule`](/api-reference/get-agent-picks-schedule) | `devtune_get_agent_picks_schedule` |
| [`PATCH /projects/{projectId}/agent-picks/schedule`](/api-reference/set-agent-picks-schedule) | `devtune_set_agent_picks_schedule` |

## Measurements

Use the summary for install rates, likely ranges and coverage. Check whether it describes your project's test or a public category test. Use starter-test progress for live individual picks and the paid-test endpoint for credit reservation and settlement.

The run list and run detail expose your project's retained observations, with the same withheld text as the app. They do not expose the source runs of a copied public test. Follow the pagination fields to read the remaining records.

The Actions endpoint includes loss evidence, re-test status, credits and results. The shared Actions and Results APIs also support the `ai_search`, `agent_picks` and `both` source selections.

## Changes

The settings endpoint changes the agent/model selections of an existing setup. It does not create the setup or start a test.

Read the schedule estimate before changing its cadence. Send its current `fingerprint` when enabling a recurring schedule. Turning the schedule off stops future tests; it does not cancel one already running.

Starting or cancelling a test, and confirming a paid re-test, happen in the app. REST and MCP reads do not start tests.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.