The combination that earns its place is timeline plus log download: the timeline names the stages, jobs and tasks and the agent each one ran on, and you can then pull that job's log by name rather than hunting for an ID. Downloads are tracked, so the temporary directory can be listed and cleaned rather than quietly filling up. Start with the health check — it reports which permissions your token actually carries, which is where most setups go wrong.
An Azure DevOps server aimed at builds and the agents behind them: pipelines and builds, agent pools and queues, and the logs and artifacts a run produced.
- Builds listed and filtered by pipeline name, status, result, branch or date range, with paging for large result sets
- Pipeline definitions listed by name, which is how you find the definition ID other calls need
- A build's timeline: every stage, job and task, and which agent executed each one
- New builds queued against a definition, with a source branch and parameters
- Job logs downloaded by job name, and logs downloaded by stage, job or task name found in the timeline — a stage pulls all its child job logs into an organized directory
- Artifacts listed for a build and Pipeline artifacts downloaded through signed URLs as a ZIP
- Agent queues in the project listed and opened, with pool details and agent counts
- Agents found across the organization's pools, or listed with filters for name, pool and online status
- The downloads this server has made: listed, located, and cleaned up by age
- A health check that verifies the connection and reports which permissions the token actually has
An Azure DevOps organization and a personal access token. Set ADO_ORGANIZATION to the organization URL, ADO_PROJECT to the project name, and ADO_PAT to the token. Scope decides what works: Agent Pools (read) and Build (read) cover the project tools, Build (read & execute) is what allows queueing a build, and the org_ tools need Agent Pools (read) at organization level because agents live there rather than in a project. Run it with npx -y @rxreyn3/azure-devops-mcp@latest over stdio, from the npm package @rxreyn3/azure-devops-mcp, currently 1.9.0. Downloads land in a managed temporary directory unless you pass an output path, and artifact download covers Pipeline artifacts — the ones published with the PublishPipelineArtifact task.
One command plus a key — claude mcp add azure-devops \ -e ADO_ORGANIZATION="https://dev.azure.com/your-organization" \ -e ADO_PROJECT="your-project-name" \ -e ADO_PAT="your-personal-access-token" \ -- npx -y @rxreyn3/azure-devops-mcp@latest, then supply credentials