Vercel MCP Server for Microsoft Copilot Studio
Enables managing Vercel projects, deployments, build logs, domains, and environment variables through natural language in Microsoft Copilot Studio, using streamable HTTP with API key authentication.
README
Vercel MCP Server for Microsoft Copilot Studio
A zero-dependency Model Context Protocol server for the Vercel platform, built specifically to work with Microsoft Copilot Studio (which requires the Streamable HTTP transport and supports API-key authentication).
Why this exists: Vercel's official MCP server (
https://mcp.vercel.com) uses OAuth with dynamic client registration, which Copilot Studio's MCP onboarding cannot complete. This server bridges the gap: the API key you enter in Copilot Studio is your Vercel access token, forwarded per-request to the Vercel REST API. The server itself is stateless and stores nothing.
Tools
| Tool | Description |
|---|---|
get_current_user |
Verify the connection / identify the token owner |
list_teams |
List Vercel teams |
list_projects |
List projects (id, framework, production URL) |
get_project |
Project details incl. git repo link |
list_deployments |
Recent deployments, filter by project/state/target |
get_deployment |
Deployment details incl. error info |
get_deployment_build_logs |
Build log output — great for diagnosing failed builds |
cancel_deployment |
Cancel a building/queued deployment |
promote_deployment |
Promote (roll forward/back) a deployment to production |
list_project_domains |
Custom domains + verification status |
list_env_vars |
Env vars (secret values hidden) |
create_env_var |
Create/upsert an env var |
delete_env_var |
Delete an env var |
Deploy
Option A — Deploy to Vercel (recommended, ~1 minute)
Or with the CLI:
vercel deploy --prod
Your MCP endpoint will be: https://<your-deployment>.vercel.app/mcp
Option B — Self-host (any Node 18+ host)
node server.js # listens on PORT (default 3000), endpoint /mcp
No npm install needed — there are no dependencies.
Connect from Copilot Studio
- Create a Vercel access token at https://vercel.com/account/settings/tokens (scope it to the team/projects you want the agent to manage).
- In Copilot Studio, open your agent → Tools → Add a tool → Model Context Protocol (MCP).
- Fill in:
- Name:
Vercel - Description:
Manage Vercel projects, deployments, build logs, domains and environment variables. - Server URL:
https://<your-deployment>.vercel.app/mcp - Authentication: API key
- Header name / location:
x-api-key(header) - Value: your Vercel access token
- Header name / location:
- Name:
- Click Add → Copilot Studio performs the MCP handshake and lists all 13 tools.
- Create connection, enable it in the connection manager, then test in the Preview pane, e.g.:
- "List my Vercel projects"
- "Show me the build logs for the latest failed deployment of my-site"
- "Add an env var API_URL=https://api.example.com to my-site for preview only"
The server also accepts the token via Authorization: Bearer <token> or x-vercel-token headers, and falls back to a VERCEL_TOKEN environment variable for single-tenant self-hosted deployments (in that case choose No authentication in Copilot Studio — only do this if the server URL itself is protected/private).
Architecture
Copilot Studio ──POST /mcp (JSON-RPC, Streamable HTTP)──▶ this server ──REST──▶ api.vercel.com
x-api-key: <vercel access token> (stateless, no storage)
- Transport: MCP Streamable HTTP, stateless mode (protocol versions 2024-11-05 → 2025-06-18). No sessions, no SSE required — ideal for serverless.
- Auth: pass-through. The server never stores tokens; every request carries the caller's own Vercel token.
- Zero dependencies: plain Node.js ≥18. Nothing to install, no supply chain.
Development & tests
node test/mcp.test.js
The test suite boots the real HTTP server, mocks api.vercel.com, and drives the full MCP handshake (initialize → initialized → tools/list → tools/call), plus error paths (bad token, unknown tool/method, malformed JSON, batching).
Security notes
- Use a scoped Vercel token, ideally on a dedicated team with least privilege.
- Secret env var values are never returned by
list_env_vars. - Deploy under HTTPS only (automatic on Vercel). Copilot Studio requires HTTPS.
- Consider Vercel Deployment Protection bypass headers if you protect the deployment.
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。