Summary
From "the pipeline failed" to "this file, this line".
The useful part is not fetching the logs — it is that the pipeline is found from the branch or PR you are already on, and only the final attempt of each step is read. What comes back is a report organised by file rather than a wall of CI output.
What it is
An MCP server for Woodpecker CI failure analysis, with two ways in: a pipeline named directly by repository id and number, or one resolved from the git context your IDE already has.
What you get
- Machine logs extracted for a given pipeline, pulled per step rather than as one undifferentiated stream
- Pipelines resolved from git context — repository name, PR number or branch — so a URL does not have to be found first
- The repository id resolved from its name, and the latest pipeline for a pull request located automatically
- Only the final attempt of a retried step analysed, so a flaky retry does not double the noise
- A structured report as Markdown and JSON, with fix suggestions organised file by file
- Two prompts shipped with it: one for direct repo-and-pipeline analysis, one for PR failures from the IDE
- Running pipelines handled without erroring out
Requirements
Two environment variables: WOODPECKER_SERVER for your Woodpecker instance and WOODPECKER_TOKEN for the account it reads as. The git-context path additionally assumes it is running where your repository is checked out.
Setup effort
One command plus a key — npm install -g woodpecker-ci-mcp, then supply credentials