# Azure DevOps MCP — Repos, Pipelines & Boards Tools

> Azure DevOps MCP turns repo, pipeline, and work-item operations into MCP tools so agents can help triage, query, and automate DevOps work safely.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

1. Install / run:
   ```bash
   git clone https://github.com/microsoft/azure-devops-mcp && cd azure-devops-mcp
   ```
2. Start / smoke test:
   ```bash
   sed -n '1,260p' README.md
   ```
3. Verify:
   - Connect the server, list available tools, then run one read-only query (pipelines/work items) before attempting any updates.

## Intro

Azure DevOps MCP turns repo, pipeline, and work-item operations into MCP tools so agents can help triage, query, and automate DevOps work safely.

- **Best for:** Teams on Azure DevOps who want agents to query pipelines and work items via structured APIs
- **Works with:** An MCP client + Azure DevOps org/project credentials (per repo docs)
- **Setup time:** 20 minutes

## Practical Notes

- Setup time ~20 minutes (auth + run server + connect agent)
- Operational check: read-only queries return within 1–2 tool calls, not multi-turn scraping
- GitHub stars + forks (verified): see Source & Thanks

DevOps automation fails when it’s ‘too powerful by default’. Keep your agent on a tight leash: start with read-only triage, then add one write action at a time. If the agent can’t provide a justification (link, run id, work item id), block the write.

### FAQ

**Q: Why MCP instead of REST calls in the agent?**
A: MCP gives you a consistent tool contract and a place to enforce auth, logging, and limits.

**Q: What should I allow first?**
A: Start with read-only tools: list pipelines, get run status, list work items.

**Q: How do I avoid risky changes?**
A: Gate write tools behind approvals and limit scopes to one project/team at a time.

## Source & Thanks

> Source: https://github.com/microsoft/azure-devops-mcp
> License: MIT
> GitHub stars: 1,669 · forks: 548

---

<!-- ZH -->

## 快速使用

1. 安装 / 运行：
   ```bash
   git clone https://github.com/microsoft/azure-devops-mcp && cd azure-devops-mcp
   ```
2. 启动 / 冒烟测试：
   ```bash
   sed -n '1,260p' README.md
   ```
3. 验证：
   - 连接 server 后先列出工具，再做一次只读查询（流水线/工作项），确认无误再尝试更新类操作。

## 简介

Azure DevOps MCP 把仓库、流水线与工作项等能力封装成 MCP 工具，让 agent 做查询、分流与自动化，同时仍能用权限范围、审批与日志审计把风险锁住，适合把日常运维动作结构化交给 agent 执行。

- **适合谁：** 使用 Azure DevOps 的团队：希望 agent 通过结构化 API 查询流水线与工作项
- **可搭配：** 支持 MCP 的客户端 + Azure DevOps 组织/项目凭证（以仓库文档为准）
- **准备时间：** 20 分钟

## 实战建议

- 上手约 20 分钟（鉴权 + 启动 server + 连接 agent）
- 效率验收：只读查询应在 1–2 次工具调用内返回，而不是多轮“爬页面”
- GitHub stars + forks（已核验）：见「来源与感谢」

DevOps 自动化出问题，往往是因为“默认太强”。让 agent 先只读分流，再逐个开放写入动作；凡是不能给出明确依据（链接、run id、work item id）的写入请求，一律拦截。

### FAQ

**为什么不让 agent 直接调 REST？**
A: MCP 把工具契约统一起来，同时方便在 server 侧做鉴权、审计与限流。

**第一步应该开放哪些能力？**
A: 优先只读：列流水线、查运行状态、列工作项。

**如何避免危险改动？**
A: 写入类工具走审批，并把 scope 限制在单个项目/团队。

## 来源与感谢

> Source: https://github.com/microsoft/azure-devops-mcp
> License: MIT
> GitHub stars: 1,669 · forks: 548


---
Source: https://tokrepo.com/en/workflows/azure-devops-mcp-repos-pipelines-boards-tools
Author: MCP Hub