CookiesForAI

Vercel MCP server

The Vercel MCP server gives AI agents access to your Vercel account through Vercel's hosted endpoint: search the docs, inspect projects and deployments, read runtime logs and errors, and query Web Analytics.

By Vercel Official · Updated

Connect Vercel

All detected agents at once

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

Claude Code (then type /mcp to sign in)

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

Codex CLI

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

Cursor .cursor/mcp.json

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

Adding the server does not finish setup: sign in to Vercel in the browser and grant the connection access to each team whose projects the agent should see. Claude.ai and ChatGPT add it as a custom connector with the URL https://mcp.vercel.com, on plans that allow custom connectors.

What the Vercel MCP server does

The Vercel MCP server is Vercel's own hosted server for connecting AI tools to a Vercel account, in public beta on 2026-10-01. After OAuth sign-in and granting access to a team, an agent can list projects and deployments, inspect a deployment's status, read and filter runtime logs, group runtime errors by route, fetch protected preview URLs you can access, and query visitors, page views and custom events. Docs search works without signing in.

Vercel's tools reference listed 213 tools in 28 categories on 2026-10-01, well beyond reading: domains and DNS, environment variables, feature flags, firewall rules, rolling releases, sandboxes, and billing and purchases. Vercel's changelog says the server can now deploy code and make purchases, so a connected agent can change production settings and spend money within your permissions.

Vercel only allows AI clients it has reviewed to connect, and requires explicit consent for each client connection to block confused deputy attacks. Its docs list 14 supported clients, and the add-mcp command also configures others such as Cline, Kilo Code and Zed. The server has no public source repo; it is listed in the MCP Registry as com.vercel/vercel-mcp.

Tools it gives your agent

  • list_deployments
  • get_deployment
  • web_fetch_vercel_url
  • get_runtime_logs
  • get_runtime_errors
  • list_projects
  • get_project
  • list_teams

When to use Vercel

  • A deployment failed and you want the agent to read the build and runtime logs.
  • You want to know which routes are throwing errors in production.
  • You want the agent to check a protected preview URL after a push.
  • You want traffic numbers for a page without opening the dashboard.

When to pick something else

  • Agents without confirmation steps: the server has the same access as your Vercel account, including deploys, environment variables and purchases, so keep human approval on.
  • Unlisted clients: only clients Vercel has reviewed can connect.

What Vercel needs

  • A Vercel account, signed in through OAuth with access granted to the team you want
  • A supported MCP client

What the Vercel MCP server costs

Vercel's docs list no separate charge for the server, which is in public beta; anything it deploys or buys is billed under your Vercel plan.

Current prices: vercel.com . We do not list prices: vendors change them often, so check the publisher's own page.

  • AWS: The AWS MCP servers give AI agents access to AWS: a managed AWS MCP Server for API calls and docs, plus more than 50 open-source servers for services like CDK, EKS, Lambda and DynamoDB.
  • Cloudflare: The Cloudflare MCP servers give AI agents access to a Cloudflare account through hosted endpoints: read Workers logs and builds, manage bindings, check DNS analytics, search the docs and render web pages.
  • Kubernetes: The Kubernetes MCP server gives AI agents direct access to Kubernetes and OpenShift clusters: list and edit any resource, read pod logs, exec into pods, view events and install Helm charts.
  • Sentry: The Sentry MCP server gives AI coding agents access to your Sentry account: search errors and events, read issue details and stack traces, look at performance data and triage issues while fixing code.