mcp-server
An MCP server that provides tools for Trunk CI Autopilot to apply fixes to failing tests.
Deployment
https://mcp.trunk.io/mcpMCP Tools (3/3)
Helps users set up flaky test uploads to Trunk. Returns instructions that must be followed and implemented. If no test framework is provided, use detect-frameworks tool to detect frameworks in your codebase, then call this tool again with the specific framework. Only one framework can be set up at a time. If org slug is not provided, it will be looked up from the user's org membership.
Returns instructions to analyze the current codebase to detect which test frameworks are present. Follow and implement the returned instructions to analyze the codebase in preparation for setting up trunk uploads.
Helps users fix CI failures. Use this when a user wants to retrieve root cause analyses and fix recommendations for CI failures. If a user wants to fix failing tests, first retrieve the following: repo name (if you don't have access to git tooling run `git remote -v` and look for the repo name in the format of "owner/repo"), (Optional) trunk.io org slug, only if the user specifies it, (Optional) Fix ID, only if the user specifies it, (Optional) PR Review comment URL, only if the user specifies it. This allows users to query older analyses directly. (Optional) Branch, if Fix ID and PR Review comment URL are not provided (if you don't have access to git tooling run `git rev-parse --abbrev-ref HEAD`). Then call this tool with those arguments. Returned results will indicate previously run root cause analyses for CI jobs that have failed tests, including the branch that the analysis was run on. Make sure that you are on the correct branch before applying the fix recommendations. If no results are returned, either the arguments are wrong or the user has not run the tests in CI and they will need to be run locally.