What it does
The Azure DevOps MCP server connects AI agents directly to your organization's ADO resources, enabling Claude and other language models to query and manipulate your development infrastructure. It exposes tools to list and interact with projects, builds, repositories, work items, test plans, iterations, teams, and wikis. The server supports both reading ADO state (listing items, retrieving content) and writing operations (creating and updating wiki pages, managing work items). Microsoft offers two deployment models: a remote HTTP endpoint (recommended for most users and in public preview) that will eventually replace the local implementation, and a local stdio server for scenarios requiring self-hosted execution.
Who it's for
Development teams and individual engineers using Azure DevOps who want to integrate AI capabilities into their workflows. DevOps engineers automating infrastructure queries, technical leads managing work-item triage, and full-stack developers seeking AI-assisted navigation of their repositories represent the core audience.
Common use cases
- Query ADO projects, repositories, and build pipelines from Claude Code
- List work items assigned to you or scheduled in the active iteration
- Retrieve test plans, team members, and project iterations programmatically
- Create and update wiki pages without switching to the web UI
- Ground AI prompts in real ADO context for more informed code and process suggestions
Setup pitfalls
- A secret was detected in the codebase—immediately review and rotate any exposed ADO credentials, storing only in environment variables
- The local
stdioserver requires Node.js 20+ and your organization name as a startup argument; Microsoft recommends the Remote MCP Server for most scenarios - ADO authentication (personal access token, alternate credentials, or managed identity) must be treated as secrets with proper access controls and never hardcoded
- The server's broad filesystem and network access, combined with its high-risk classification, warrants isolation in a container or restricted environment