Netlify’s official MCP server closes the loop between an agent writing code and that code being live. It can create a site, deploy it, read the deploy logs when the build fails, set environment variables, install extensions and manage forms — the whole path from local directory to production URL.
Bring your own
Claude Sonnet 4.5
Chat with 60+ AI models on the same workflow — switch to a different model mid-conversation and re-run the same prompt, or use Compare mode to put several side-by-side and balance quality vs. cost.
Your existing Netlify CLI login, or a personal access token in NETLIFY_PERSONAL_ACCESS_TOKEN when you want to scope it explicitly.
How models use it and what it is built for.
Most deploy integrations stop at "push and hope". This one gives an agent the feedback it needs to actually finish the job: after a deploy it can read the build logs, see why the build failed, fix the configuration and redeploy. Around that core it covers the platform surface — creating and listing sites, managing environment variables per context, installing and configuring extensions such as Auth0 or a database provider with preconfigured defaults, reading form submissions, and handling team and user management. Because it runs over stdio through the Netlify CLI package, it authenticates with whatever Netlify session you already have on the machine, or with a personal access token when you want to scope it explicitly. It works with any MCP-capable coding agent — Claude Code, Cursor, Windsurf, VS Code Copilot.
Typical tools an AI model can call. Exact names vary by version.
Taken from the official Netlify documentation — see netlify/netlify-mcp — official repository for the full reference.
Environment variables
NETLIFY_PERSONAL_ACCESS_TOKENPersonal access token from Netlify user settings. Optional if you are already logged in with the Netlify CLI.
Client configuration
npx (recommended)
Uses the Netlify session already on the machine — no token needed if you have run netlify login.
{
"mcpServers": {
"netlify": {
"command": "npx",
"args": ["-y", "@netlify/mcp"]
}
}
}With an explicit access token
Use this in CI, or when you want the agent scoped to a specific account rather than your CLI session.
{
"mcpServers": {
"netlify": {
"command": "npx",
"args": ["-y", "@netlify/mcp"],
"env": { "NETLIFY_PERSONAL_ACCESS_TOKEN": "YOUR_TOKEN" }
}
}
}Copy any of these into MCP Agent Studio after connecting.
Deploy this project to a new Netlify site and give me the URL.
The last deploy failed — read the build log and tell me what broke.
Set these environment variables on the production context only.
Which of my sites have not been deployed in the last 90 days?
This is not a single-model product: you get the same MCP connection with 60+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.
Default pick for Netlify
Claude Sonnet 4.5
The valuable loop here is deploy, read the failing build log, fix, redeploy. Sonnet 4.5 diagnoses build errors accurately enough to make that loop terminate.
Listing tools proves the server is reachable, not that a model can work with it. Evals go further: they read every tool on the server, write a test suite from its real schemas, and run it — code decides pass/fail on the responses (schema conformance, error codes, pagination, result caps) while a scoring model grades plain-English tasks driven through the tools.
Get a pass/fail report per tool with the evidence behind each verdict — and replay the same suite after every schema change. Destructive tools are excluded from the run.
Run evalsOpen MCP Agent Studio with the connection pre-filled. Add your token, pick any of 60+ models, and start chatting — no install required.
Open Agent StudioCommon questions about connecting, scoping and using it safely.
It is Netlify’s official MCP server, published as @netlify/mcp. It lets an AI coding agent create and deploy sites, read deploy logs, manage environment variables, install extensions and handle forms and team settings.
It can, which is exactly why you should decide deliberately. Pointing the agent at a branch deploy or a preview context gives you the same feedback loop without a production change, and is the sane default while you are learning how it behaves.
It uses your existing Netlify CLI session if you have one, so netlify login is often the whole setup. Setting NETLIFY_PERSONAL_ACCESS_TOKEN instead is the better route for CI or when you want the agent bound to one specific account.
Because it is what turns a one-shot deploy into a loop that converges. Without log access the agent deploys, the build fails, and it has nothing to act on. With it, the agent reads the actual error, fixes the config and redeploys — usually in one or two rounds.
Yes. It is a standard stdio MCP server and works with Cursor, Windsurf, VS Code Copilot, Claude Code and anything else that speaks MCP. You can also test it against any model in MCP Agent Studio before committing to one.
Vercel
List deployments, read logs, manage env keys and roll back from natural language.
Cloudflare
Drive Workers, DNS, R2, D1 and the rest of the Cloudflare API with AI.
GitHub
Drive GitHub repos, PRs and issues with an AI agent.
Docker
List containers, inspect images and manage your Docker environment with AI.