Vercel MCP Server: Tools, Setup and Security

Vercel MCP lets AI tools search Vercel docs, manage projects and deployments, read deployment logs and query Web Analytics. Vercel calls it "Vercel's official MCP server," a remote server with OAuth at https://mcp.vercel.com.

Checked against: Vercel MCP documentation on 28 October 2026. eSEOspace did not install or test this server.

At a glance

What it can do

Vercel's tools reference says it covers "217 catalog tools and 31 workflow or widget tools." What you see depends on server configuration, your client, and team or project scope.

Vercel lists these as frequently used:

• list_teams and list_projects to find your resources.
• get_project to inspect a project's settings.
• list_deployments and get_deployment to check deployment status.
• get_runtime_logs and get_runtime_errors to troubleshoot.
• web_fetch_vercel_url to fetch a deployment URL, including protected deployments you can access.

Tools are grouped into categories. Larger ones include Domains and DNS (28 tools), Sandboxes (23), Deployments (17), Projects (17), Routing (17), Feature Flags (16) and Vercel CI (15). Others cover environment variables, firewall settings, rolling releases, Web Analytics, and billing and purchases.

Vercel's changelog, linked from the docs, notes that Vercel MCP "can now deploy code," "now supports purchases" and supports the 2026-07-28 MCP specification.

Some actions are left out on purpose. Vercel says the catalog does not create AI Gateway API keys, exchange SSO tokens, import KMS private keys, or sign KMS messages or tokens.

How to connect it

Vercel's quick setup configures every detected client at once:

npx -y add-mcp https://mcp.vercel.com -g

For Claude Code, Vercel documents:

claude mcp add --transport http vercel https://mcp.vercel.com

Then type /mcp inside Claude Code to sign in. For Cursor, add this to .cursor/mcp.json:

{
  "mcpServers": {
    "vercel": {
      "url": "https://mcp.vercel.com"
    }
  }
}

Vercel's supported list also includes Claude.ai and Claude for desktop, ChatGPT, Codex CLI, Devin, Raycast, Goose, Windsurf (now Devin Desktop), Gemini Code Assist, Gemini CLI, OpenCode and Pi.

Adding the server does not finish authorization. You must sign in and grant access to each team you want the AI to work with. Your normal Vercel permissions still apply.

Security notes

Vercel's docs list these practices:

• Verify the endpoint. Always confirm you are connecting to https://mcp.vercel.com, especially with third-party "one-click" installers.
• Know the scope. Vercel says connecting "grants the AI system you're using the same access as your Vercel user account."
• Confused deputy protection. Vercel requires explicit consent for each client connection. See the confused deputy problem in MCP.
• Prompt injection. Vercel warns that untrusted tools could instruct an agent to copy private deployment logs elsewhere.
• Human confirmation. Vercel says to "always enable human confirmation in your workflows."

Separate auth is another point to watch. Vercel notes MCP and the Vercel CLI authenticate separately, so a CLI success does not prove your MCP connection has the same access. After an uncertain write, inspect the resource before retrying. More guidance is in the MCP security hub.

Alternatives

• Netlify MCP Server for sites hosted on Netlify.
• Cloudflare MCP Server for Workers, Pages and DNS.
• Render MCP Server for web services, cron jobs and Postgres on Render.

Next step

For website teams on Vercel, start by granting one team and asking the AI to list your projects. Our list of MCP servers for developers compares hosting servers side by side. If you want help building agent workflows around deployments, see our Claude Code development service.

Frequently asked questions