ado-mcp
Enables AI assistants to interact with Azure DevOps, providing tools for managing work items, repositories, pipelines, and more through the Model Context Protocol.
README
Azure DevOps MCP Server
A production-grade Model Context Protocol server for Azure DevOps. Exposes Azure DevOps operations as MCP tools (plus, in later phases, resources and prompts) over stdio (local) and Streamable HTTP (hosted/multi-user).
- TypeScript (ESM, Node 20+)
- ADO access via the official
azure-devops-node-apiSDK, with an axios REST fallback for endpoints the SDK doesn't cover (search, analytics, etc.) - Three auth modes: PAT, Azure CLI (
az login), On-Behalf-Of + Entra JWT (hosted) - Per-tool RBAC, destructive-op guard, secret redaction, structured logging, retry with backoff
Status: All 6 phases complete — 113 tools across 13 domains, plus 4 MCP resources and 4 MCP prompts:
- core (7) · repos (15) · pullrequests (12) · workitems (16) · boards (10)
- pipelines (13) · releases (8) · testplans (10) · wiki (7) · search (3) · artifacts (5)
- security (5) · analytics (2)
- Resources:
ado://projects,ado://project/{project}/repos,ado://project/{project}/workitem/{id},ado://project/{project}/repo/{repo}/pullrequest/{prId}- Prompts:
pr-review,release-readiness,sprint-planning,bug-triageHardening: per-host circuit breaker, retry w/ backoff + Retry-After, secret redaction, per-tool RBAC + destructive guard, audit logging, ESLint, a Vitest suite (unit + live MCP-protocol test), and an Azure Pipelines CI (
azure-pipelines.yml).
1. Prerequisites
- Node.js ≥ 20 and npm
- An Azure DevOps organization you can access
- One of:
- a Personal Access Token (PAT), or
- the Azure CLI installed and
az logincompleted, or - (hosted) an Entra app registration with the ADO delegated permission, for OBO
2. Install
cd "E:\MCP Servers\ado-mcp-server"
npm install
3. Configure
cp .env.example .env # PowerShell: Copy-Item .env.example .env
Edit .env. Minimum for a quick local start with the Azure CLI:
ADO_ORG=YourOrgName
ADO_DEFAULT_PROJECT=YourProject
AUTH_MODE=azcli
Auth modes
AUTH_MODE |
Needs | Notes |
|---|---|---|
azcli |
az login done locally |
Easiest for dev. No secrets stored. |
pat |
ADO_PAT |
PAT scopes must cover the tools you call (e.g. Work Items: Read & Write, Code: Read & Write). |
obo |
AZURE_TENANT_ID, AZURE_CLIENT_ID, AZURE_CLIENT_SECRET |
HTTP transport only. Validates the caller's Entra JWT, then exchanges it for an ADO token preserving the user's identity. |
A PAT is created in Azure DevOps → User settings → Personal access tokens.
4. Build & run
npm run build # compile to dist/
npm start # stdio transport (default)
Development with auto-reload:
npm run dev # stdio
npm run dev:http # Streamable HTTP on PORT (default 3000)
Force a transport regardless of TRANSPORT:
node dist/index.js --transport http
node dist/index.js --transport stdio
5. Connect a client
Claude Desktop / VS Code / Claude Code (stdio)
Add to your MCP client config (e.g. Claude Desktop claude_desktop_config.json).
Published package via npx (recommended distribution — no clone, no build):
{
"mcpServers": {
"azure-devops": {
"command": "npx",
"args": ["-y", "ado-mcp-server"],
"env": {
"ADO_ORG": "YourOrgName",
"ADO_DEFAULT_PROJECT": "YourProject",
"AUTH_MODE": "pat",
"ADO_PAT": "<your-pat>"
}
}
}
}
Local build via node (for development on a clone):
{
"mcpServers": {
"azure-devops": {
"command": "node",
"args": ["E:\\MCP Servers\\ado-mcp-server\\dist\\index.js"],
"env": {
"ADO_ORG": "YourOrgName",
"ADO_DEFAULT_PROJECT": "YourProject",
"AUTH_MODE": "azcli"
}
}
}
}
The org is never hardcoded — it comes from
ADO_ORGin the client'senvblock, so the same published package serves any org.
Remote / hosted (Streamable HTTP)
node dist/index.js --transport http
Endpoint: POST http://localhost:3000/mcp · health: GET /health.
In obo mode every request must carry Authorization: Bearer <Entra access token>.
6. Verify it works
# Health (HTTP mode)
curl http://localhost:3000/health
# From an MCP client, call the connectivity probe tool:
# ado_get_me → confirms auth + org connection
# ado_list_projects
A scripted stdio smoke test (lists tools without calling ADO):
printf '%s\n' \
'{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"smoke","version":"1.0"}}}' \
'{"jsonrpc":"2.0","method":"notifications/initialized"}' \
'{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}' \
| ADO_ORG=x AUTH_MODE=pat ADO_PAT=dummy node dist/index.js
7. Publishing (npx distribution)
This server is distributed as an npm package so anyone can run it with
npx ado-mcp-server — no clone or build on their side.
npm run build # compile (also runs automatically via prepublishOnly)
npm version patch # bump version
npm publish # publishes; only dist/, README.md, .env.example ship (see "files")
After publishing, the npx client config in §5 works as-is. To test the
package contents before publishing:
npm pack --dry-run # lists exactly what will be uploaded
npm pack # builds the .tgz locally; install it to smoke-test
Auth for distributed use: prefer
AUTH_MODE=pat— each user supplies their ownADO_PAT(andADO_ORG) in their MCP client'senv.azclionly works where the user has the Azure CLI installed andaz logincompleted.
8. Scripts
| Script | Purpose |
|---|---|
npm run build |
Compile TypeScript → dist/ |
npm start |
Run compiled server (stdio) |
npm run start:http |
Run compiled server (HTTP) |
npm run dev / dev:http |
Watch-mode dev (tsx) |
npm run typecheck |
tsc --noEmit |
npm run lint |
ESLint |
npm test |
Vitest |
9. Configuration reference
See .env.example. Key vars: ADO_ORG, ADO_DEFAULT_PROJECT, ADO_API_VERSION, AUTH_MODE, ADO_PAT, AZURE_TENANT_ID/CLIENT_ID/CLIENT_SECRET, TRANSPORT, PORT, LOCAL_ROLE, LOG_LEVEL, ALLOW_DESTRUCTIVE.
ALLOW_DESTRUCTIVE— destructive tools (delete_*, complete PR, deploy release, …) are blocked unless this istrue.LOCAL_ROLE— effective RBAC role inpat/azclimodes (admin/editor/viewer). Inobomode roles come from Entra App Roles on the caller's token.
10. Architecture (short)
transport (stdio | http)
→ server (McpServer + registry)
→ dispatch (RBAC · destructive guard · audit · error envelope)
→ domain tools (Zod schemas)
→ domain services
→ AdoClient (azure-devops-node-api SDK | REST fallback + retry)
→ auth (PAT | AzCli | OBO credential)
Each ADO area is a self-contained module under src/domains/<area>/ exporting a DomainModule. To add an area, create the module and append it to domains in src/server.ts — nothing else changes.
Full walkthrough: see docs/END_TO_END_FLOW.md for how a request travels through every layer (startup, a tool call step by step, auth flows, resources/prompts, reliability, and a worked example).
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.