@circleci/mcp-server-circleci

What it does

This MCP server bridges CircleCI into LLM clients—Claude, Cursor, VS Code, and others—exposing CI/CD operations as natural-language tools. It can trigger and rerun workflows, fetch test results and failure logs, identify flaky tests across runs, download usage and resource data for cost optimization, and validate CircleCI configurations, all without leaving your editor.

Who it's for

Platform engineers and DevOps teams automating CI/CD workflows from their IDE; backend engineers debugging build failures and test flakiness; teams analyzing CircleCI costs and resource allocation to optimize spending.

Common use cases

• Trigger a pipeline and check its status directly from your editor without browser switching.
• Analyze flaky test patterns across multiple runs to identify stability regressions.
• Download CircleCI usage metrics and resource-class data to optimize job provisioning and reduce costs.
• Retrieve detailed build failure logs and job output to debug issues faster.
• Validate CircleCI configuration syntax and catch errors before committing.

Setup pitfalls

• Requires a CircleCI Personal API token; create one at app.circleci.com/settings/user/tokens and pass it via the CIRCLECI_TOKEN environment variable at startup.
• On-premises CircleCI customers must set CIRCLECI_BASE_URL explicitly; it defaults to the cloud SaaS endpoint.
• Filesystem write access required if using download_usage_api_data; ensure the process can write to your target directory.