Home

Welcome to the Azure DevOps MCP Server Wiki

This wiki contains detailed documentation and reference guides for the Azure DevOps Model Context Protocol (MCP) server.

Caution

CRITICAL WARNING: Destructive Actions & Work Item Deletions

This MCP server is powerful and exposes tools (and a generic REST client) that allow an AI assistant to perform any API call in your Azure DevOps organization (such as committing/pushing code, deleting repositories, triggering builds, or altering work items).
Azure DevOps does NOT keep a recycle bin or trashcan for work items deleted via the REST API. If a work item (e.g., Bug, Task, User Story) is deleted via the API (using api_call or other means), it is permanently destroyed and cannot be restored.
Always review and approve suggested operations carefully, especially when granting write/delete permissions or confirming commands.

Explore the detailed sections of the documentation:

Learn how to generate a Personal Access Token (PAT) with the correct scopes.
Understand the credential security architecture (AES-256-GCM local encryption).
Install via Smithery or manually, and integrate the server with AI clients like Claude Desktop or Antigravity.

Run the server over HTTP (Streamable HTTP at /mcp, legacy SSE at /sse), locally or in Docker.
Protect the endpoint with a bearer token or Microsoft Entra ID, and pass per-session Azure DevOps credentials.

Complete specifications for all 23 exposed tools.
Parameter lists, required fields, and examples for Work Item Tracking (WIT), Git Integration, Pipeline Management, and Identity Search.

How to use the powerful api_call tool.
Explaining the offline API directory (api_docs and api_info) which helps AI models discover and execute arbitrary Azure DevOps REST API requests.

How to run the automated unit tests.
Creating a free sandbox Azure DevOps organization and populating test data to verify the MCP tools safely.