Vercel MCP Server

Official remote MCP server from Vercel that lets AI tools search docs, manage projects and deployments, view runtime logs, and handle domains.

Developer Tools by Vercel OAuth2 active

Connect (Remote Server) Docs

Overview

Vercel MCP is Vercel's official remote Model Context Protocol server, available at https://mcp.vercel.com. It exposes Vercel platform capabilities to AI assistants over an OAuth-secured Streamable HTTP transport, giving agents the same scope of access as the authorized user's Vercel account. Available in Beta on all plans, it is designed to be the canonical way for AI tools to interact with Vercel without managing personal API tokens.

The server provides tools across several domains: documentation search, team and project management, deployment inspection (including build logs and runtime function logs), domain availability checks and purchasing, deployment protection bypass via shareable links, fetching content from protected deployment URLs, and reading and replying to Vercel Toolbar comment threads. It also includes helper tools that prompt the LLM to use the Vercel CLI for deploys and other actions.

Vercel MCP only supports AI clients that have been explicitly reviewed and approved by Vercel, including Claude Code, Claude.ai, Claude Desktop, ChatGPT, Codex CLI, Cursor, VS Code with Copilot, Devin, Raycast, Goose, Windsurf, and the Gemini Code Assist and Gemini CLI tools. It implements the MCP Authorization spec (2025-06-18) and includes confused deputy protection requiring explicit user consent per client connection.

Tools

Tool	Description
`search_documentation`	Search Vercel documentation for specific topics and information.
`list_teams`	List all Vercel teams that include the authenticated user as a member.
`list_projects`	List all Vercel projects associated with a user under a given team.
`get_project`	Get detailed information about a specific project including framework, domains, and latest deployment.
`list_deployments`	List deployments for a project with creation time, state, and target information.
`get_deployment`	Get detailed information for a specific deployment including build status, regions, and metadata.
`get_deployment_build_logs`	Get the build logs of a deployment by ID or URL to investigate failures.
`get_runtime_logs`	Get runtime logs for a project or deployment, filterable by environment, level, status code, source, time range, and full-text search.
`check_domain_availability_and_price`	Check if domain names are available for purchase and get pricing information.
`buy_domain`	Purchase a domain name with registrant information.
`get_access_to_vercel_url`	Create a temporary shareable link that grants access to protected Vercel deployments.
`web_fetch_vercel_url`	Fetch content directly from a Vercel deployment URL, with authentication if required.
`list_toolbar_threads`	List Vercel Toolbar comment threads for a team. Returns unresolved threads by default.
`get_toolbar_thread`	Get a specific toolbar thread by ID, including all messages and context.
`change_toolbar_thread_resolve_status`	Mark a toolbar thread as resolved or unresolve it.
`reply_to_toolbar_thread`	Add a reply message to an existing toolbar thread.
`edit_toolbar_message`	Edit an existing message in a toolbar thread.
`add_toolbar_reaction`	Add an emoji reaction to a message in a toolbar thread.
`use_vercel_cli`	Instructs the LLM to use Vercel CLI commands (with --help) to accomplish a task.
`deploy_to_vercel`	Deploy the current project to Vercel.

Setup Guide

Vercel MCP is a remote, OAuth-authenticated MCP server. There is nothing to install or self-host. You connect your AI client to https://mcp.vercel.com and complete the OAuth flow in the browser to authorize access to your Vercel account.

Quick install (auto-detect clients)

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

Add -y to skip the confirmation prompt, or -g to install globally across all projects.

Claude Code

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

Then run claude and type /mcp to authenticate.

Claude.ai or Claude Desktop (Pro/Max/Team/Enterprise)

In Settings → Connectors → Add custom connector, set:

Name: Vercel
URL: https://mcp.vercel.com

Cursor

Add to .cursor/mcp.json (project or global):

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

Windsurf

Add to mcp_config.json:

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

Gemini Code Assist / Gemini CLI

Add to ~/.gemini/settings.json (uses mcp-remote bridge):

{
  "mcpServers": {
    "vercel": {
      "command": "npx",
      "args": ["mcp-remote", "https://mcp.vercel.com"]
    }
  }
}

ChatGPT (Pro/Plus, Developer mode)

In Settings → Connectors, create a new connector with name Vercel, MCP server URL https://mcp.vercel.com, and Authentication OAuth.

Codex CLI

codex mcp add vercel --url https://mcp.vercel.com

Prerequisites

A Vercel account
An approved MCP client from the supported list
For Claude.ai connectors, a Pro/Max/Team/Enterprise plan
For ChatGPT connectors, a Pro/Plus account with Developer mode enabled

Use Cases

Debug a failing deployment by pulling the build logs with get_deployment_build_logs and the runtime function logs with get_runtime_logs, then ask the agent to propose a fix.
Triage production incidents by filtering runtime logs by level: error and statusCode: 5xx over the last hour to surface the offending request and stack trace.
Look up Vercel documentation in-context with search_documentation while wiring up a new feature, without leaving the IDE.
Inspect a team's projects and the latest production deployment per project to generate a release summary or status report.
Manage post-launch QA from chat: list unresolved Vercel Toolbar comment threads on a project, reply to them, and mark them resolved once fixed.
Check availability and price for a list of candidate domain names before launch, and purchase the chosen one.

Example Prompts

"Show me the build logs for the latest failed deployment of my marketing-site project and tell me what broke."
"Get runtime error logs from the last hour for project prj_abc in production, grouped by status code."
"How do I configure custom domains in Vercel?"
"List unresolved Vercel Toolbar comments on the blog project and draft replies for each."
"Check if acme-launch.com and acme-launch.app are available, and tell me the prices."
"Create a shareable link for https://my-app-preview.vercel.app so I can send it to a contractor."

Pros

Official, Vercel-maintained remote MCP server with OAuth, so no API token management or self-hosting required.
Broad tool coverage across projects, deployments, build and runtime logs, domains, Toolbar comments, and docs search.
Implements the latest MCP Authorization and Streamable HTTP specs, with confused deputy protection requiring per-client consent.
Works out of the box with most major MCP clients (Claude, ChatGPT, Cursor, VS Code, Codex, Gemini, Windsurf, Raycast, Devin, Goose).

Limitations

Restricted to an allowlist of Vercel-approved AI clients; arbitrary or homegrown MCP clients cannot connect.
No direct tool for managing environment variables, secrets, or project settings; mutations are largely limited to domains, Toolbar threads, and deploys via the CLI helper.
The deploy and CLI helpers route through Vercel CLI on the user's machine rather than executing remotely, so they require a local Vercel CLI install.
Still in Beta, so the tool surface and behavior may change.

Alternatives

Vercel CLI (vercel) invoked directly by an agent's shell tool, for self-managed token auth and full command coverage.
Community Vercel MCP implementations on GitHub (e.g. nganiet/mcp-vercel) for self-hosted setups using a personal Vercel API token.
Netlify MCP server or Cloudflare MCP servers for similar capabilities on competing hosting platforms.