buildkite-mcp-server — Claude Code MCP server: features & install
Exposes Buildkite CI/CD pipelines to your AI coding assistant. Helps investigate broken runs, fetch tail logs, compare builds with main, and trigger safe step retries.
claude mcp add --env BUILDKITE_API_TOKEN=<BUILDKITE_API_TOKEN> --transport stdio buildkite-mcp-server -- docker run -i --rm -e BUILDKITE_API_TOKEN ghcr.io/buildkite/buildkite-mcp-server:0.7.0What buildkite-mcp-server does
The server links your coding assistant to the Buildkite API over stdio or HTTP. Instead of pasting console logs back and forth, you let the model inspect pipeline failures directly. It queries build statuses and reads tail log outputs directly inside your workspace.
Available tools are split into modular sets. The retries toolset handles failure investigations and lets the model rerun infrastructure failures without broad write access. You only need read_builds, read_build_logs, and retry_builds scopes instead of write_builds. It also includes a dry_run option to verify proposed job retries before touching anything.
The compare_builds helper pinpoints regressions by matching a failed run against an earlier successful build on the same branch. It tracks up to 100 job transitions and pulls the last log lines for critical blockers. Running this server requires Docker and a valid BUILDKITE_API_TOKEN.
What it helps with
- find pipeline failure root causes from recent job logs
- compare broken builds against the last passing run on the branch
- trigger infrastructure retries safely after dry-run review
- inspect test suite monitoring rules and alert workflows
Example prompts for the agent
- Check build 89 in pipeline web-app and show logs for failed jobs.
- Compare build 142 with the last passing run on main in the api pipeline.
- Run a dry-run retry for infrastructure failures on build 210.
How to install buildkite-mcp-server
Platform CLI
claude mcp add --env BUILDKITE_API_TOKEN=<BUILDKITE_API_TOKEN> --transport stdio buildkite-mcp-server -- docker run -i --rm -e BUILDKITE_API_TOKEN ghcr.io/buildkite/buildkite-mcp-server:0.7.0Replace <…> with your own values
Platform CLI
codex mcp add buildkite-mcp-server --env BUILDKITE_API_TOKEN=<BUILDKITE_API_TOKEN> -- docker run -i --rm -e BUILDKITE_API_TOKEN ghcr.io/buildkite/buildkite-mcp-server:0.7.0Replace <…> with your own values
Config file
{
"mcpServers": {
"buildkite-mcp-server": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"BUILDKITE_API_TOKEN",
"ghcr.io/buildkite/buildkite-mcp-server:0.7.0"
],
"env": {
"BUILDKITE_API_TOKEN": "<BUILDKITE_API_TOKEN>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"buildkite-mcp-server": {
"type": "stdio",
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"BUILDKITE_API_TOKEN",
"ghcr.io/buildkite/buildkite-mcp-server:0.7.0"
],
"env": {
"BUILDKITE_API_TOKEN": "<BUILDKITE_API_TOKEN>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e BUILDKITE_API_TOKEN=<BUILDKITE_API_TOKEN> buildkite-mcp-server docker run -i --rm -e BUILDKITE_API_TOKEN ghcr.io/buildkite/buildkite-mcp-server:0.7.0Replace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about buildkite-mcp-server
How to install buildkite-mcp-server in Claude Code?
Run the command `claude mcp add --env BUILDKITE_API_TOKEN=<BUILDKITE_API_TOKEN> --transport stdio buildkite-mcp-server -- docker run -i --rm -e BUILDKITE_API_TOKEN ghcr.io/buildkite/buildkite-mcp-server:0.7.0. Be sure to supply your personal <BUILDKITE_API_TOKEN>`.
What does buildkite-mcp-server do?
It provides an MCP interface for Buildkite API endpoints, allowing LLMs to inspect build pipelines, read logs, match runs against baselines, and investigate test issues.
Does retrying failed jobs require full write permissions?
No, the retries toolset works with reduced scopes. You only need read_builds, read_build_logs, and retry_builds rather than general write_builds access.
