MCP Directory

Buildkite MCP Server

Official

Official Buildkite MCP server to inspect pipelines, builds, jobs and tests and fix failed builds.

Verified
HTTP (remote)
OAuth
Go
View repo 60 Website

Add to your client

Copy the config for your MCP client and paste it into its config file.

Install / run
Connect client to https://mcp.buildkite.com/mcp and complete OAuth

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "buildkite-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.buildkite.com/mcp"
      ]
    }
  }
}

Claude Desktop connects to remote servers through the `mcp-remote` proxy (installed on first run via npx). Restart Claude Desktop after saving.

Before you start

  • A Buildkite account with access to the relevant organization and pipelines
  • For the remote server (recommended): an MCP client that supports remote HTTP MCP + OAuth — connect to https://mcp.buildkite.com/mcp (or /mcp/readonly). No token needed.
  • For the local server: a Buildkite API access token from https://buildkite.com/user/api-access-tokens with scopes such as read_builds, read_pipelines, read_user (plus read_build_logs, read_artifacts, read_clusters, read_suites for full read; write_builds, write_pipelines for writes)
  • For local install: one of Homebrew, Docker, the Go toolchain, or a pre-built binary from GitHub Releases

About Buildkite MCP Server

The Buildkite MCP Server is the official, vendor-maintained MCP integration for Buildkite CI/CD. It exposes your pipelines, builds, jobs, test runs, logs, artifacts, agents, clusters, and annotations to AI tools so an agent can inspect CI state and help you diagnose and fix failed builds.

It's offered in two flavors. The recommended remote server is a Buildkite-hosted HTTP endpoint at https://mcp.buildkite.com/mcp (with a read-only variant at /mcp/readonly) that uses OAuth — no token management on your side; tokens are short-lived (12h access, 7-day refresh). A local server (Go binary) is also available for stdio clients, authenticated with a Buildkite API access token you create yourself.

Typical workflows: an agent pulls the logs of a failed job, searches them for the error, reads the build's annotations, and proposes a fix — or triggers a rebuild, unblocks a blocked step, or retries a flaky job. Tools are grouped into selectable toolsets so you can enable only what you need and keep the remote read-only endpoint when you want a safe, inspection-only setup.

Tools & capabilities (11)

current_user / access_token / user_token_organization

Identify the authenticated user, inspect the token's scopes, and resolve the org

list_pipelines / get_pipeline / create_pipeline / update_pipeline

Browse and manage CI/CD pipelines

list_builds / get_build / create_build / cancel_build / rebuild_build

Inspect builds and trigger, cancel, or rebuild them

unblock_job / retry_job / get_job_env

Unblock blocked steps, retry failed jobs, and read job environment

search_logs / tail_logs / read_logs

Search job logs by regex, tail recent output, or read full logs

list_artifacts_for_build / list_artifacts_for_job / get_artifact

List and download build/job artifacts

list_annotations / create_annotation

Read and add build annotations

list_agents / get_agent

Inspect connected Buildkite agents

list_clusters / get_cluster / list_cluster_queues / pause_cluster_queue_dispatch

Manage clusters and queue dispatch

get_test / list_test_runs / get_test_run / get_failed_executions

Query Test Engine suites, runs, and failed executions

list_pipeline_schedules / create_pipeline_schedule

View and manage cron-based pipeline schedules

When to use it

  • Use it when a build failed and you want the agent to fetch the job logs, find the error, and suggest a fix
  • Use it when you want to trigger, cancel, or rebuild a Buildkite build from your editor or chat
  • Use it when a deploy step is blocked and you need to unblock it or retry a flaky job
  • Use it when investigating flaky tests via Test Engine runs and failed executions
  • Use it when you want a safe, read-only CI inspection endpoint for an AI agent (use the /mcp/readonly remote URL)
  • Use it when you want CI build/artifact context available to an agent without copy-pasting logs

Quick setup

  1. 1Decide between the remote (OAuth, recommended) and local (API token) server
  2. 2For remote: add https://mcp.buildkite.com/mcp (or /mcp/readonly) to your MCP client and complete the OAuth login
  3. 3For local: install via Homebrew (brew install buildkite/buildkite/buildkite-mcp-server), Docker, go install, or a release binary
  4. 4For local: create a Buildkite API token at buildkite.com/user/api-access-tokens with the needed scopes and pass it to the server
  5. 5Restart your MCP client so it loads the server, optionally enabling only the toolsets you need
  6. 6Verify by asking the agent to run current_user or list_pipelines

Security notes

The remote server exposes your organization's pipeline configuration and build logs, which can contain secrets or sensitive output; review the scopes granted during authorization. For the local variant, treat BUILDKITE_API_TOKEN as a secret.

Buildkite MCP Server FAQ

Should I use the remote or local server?

Buildkite recommends the remote server (https://mcp.buildkite.com/mcp). It uses OAuth with short-lived tokens so you don't manage API keys; the local Go binary is for stdio clients or air-gapped setups.

How do I authenticate the remote server?

Via standard OAuth handled by your MCP client — access tokens last 12 hours and refresh tokens 7 days. No manual API token is needed.

What token scopes does the local server need?

At minimum read_builds, read_pipelines, and read_user. For full read access add read_build_logs, read_artifacts, read_clusters, read_organizations, and read_suites; for write actions add write_builds and write_pipelines.

Is there a read-only option?

Yes. The remote server offers a read-only endpoint at https://mcp.buildkite.com/mcp/readonly, and you can scope local tokens to read-only and enable only read toolsets.

Is it official and open source?

Yes — it's maintained by Buildkite and released under the MIT license; the local server source is on GitHub at buildkite/buildkite-mcp-server.

#buildkite#official#ci#cd#remote#developer-tools

Alternatives to Buildkite MCP Server

Playwright MCP Server
Developer Tools
24k

Microsoft's official browser-automation MCP using Playwright's accessibility tree (no vision model).

Featured
Verified
stdio (local)
No auth
TypeScript
12 tools
Updated 9 days agoRepo
Context7 MCP Server
Developer Tools
28k

Up-to-date, version-specific library documentation injected into your coding agent.

Verified
stdio (local)
API key
TypeScript
2 tools
Updated 13 days agoRepo
Serena
Developer Tools
12k

LSP-powered coding agent toolkit: semantic symbol search, references and structural edits.

Verified
stdio (local)
No auth
Python
11 tools
Updated 11 days agoRepo

Compare Buildkite MCP Server with:

vs Playwright MCP Servervs Context7 MCP Servervs Serenavs JetBrains MCP Proxy