Vercel MCP is Vercel’s official Model Context Protocol server. It is a remote server with OAuth, hosted at https://mcp.vercel.com, so there is nothing to install or keep running yourself. Once connected, your AI tool can search and navigate the Vercel documentation, manage projects and deployments, analyze deployment logs, and query Web Analytics such as visitors, page views and custom events.
The tools split into two groups. Public tools, like documentation search, work without signing in; authenticated tools need you to authorize with your Vercel account. Vercel’s docs say the server follows the current MCP authorization and streamable HTTP specifications, and that it requires explicit user consent for each client connection as a defense against confused-deputy attacks. Connections are limited to AI clients Vercel has reviewed and approved, which currently include Claude Code, Claude, ChatGPT, Codex CLI, Cursor, VS Code with Copilot and Gemini CLI, among others.
Setup is short. npx add-mcp https://mcp.vercel.com configures it for the agents it detects, or in Claude Code you run claude mcp add --transport http vercel https://mcp.vercel.com and authenticate from /mcp. The security guidance is worth reading before you do: connecting gives the AI the same access as your Vercel user account, and Vercel recommends keeping human confirmation on so you approve each step, because untrusted content in your workflow could try to steer an agent.
Verdict: useful if you already deploy on Vercel and want debugging and deployment questions answered in your editor or chat. It is labeled beta, so expect change. For the wider picture, see the best MCP servers for developers, the MCP guide, and a similar hosted server in Supabase MCP.
Sources: Vercel’s MCP documentation, checked on September 21, 2026.
