Deploy from Claude Code, Cursor, or Windsurf

Varity TeamCore Contributors

Varity works with AI coding tools. Install the MCP server once, and your AI editor can build and deploy apps to production without you ever leaving the chat.

Works with Claude Code, Cursor, Windsurf, VS Code (Copilot), and any other editor that supports the Model Context Protocol.

Install the MCP server

Claude Code

Run this once from any directory:

claude mcp add varity -- npx -y @varity-labs/mcp

Claude Code picks up the server automatically on the next conversation. Confirm it is registered:

claude mcp list

You should see varity in the list.

Cursor

Create .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "varity": {
      "command": "npx",
      "args": ["-y", "@varity-labs/mcp"]
    }
  }
}

Restart Cursor. The Varity tools appear in the MCP tool list under the agent panel.

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "varity": {
      "command": "npx",
      "args": ["-y", "@varity-labs/mcp"]
    }
  }
}

Restart Windsurf to load the server.

VS Code (Copilot)

Add to .vscode/mcp.json in your workspace:

{
  "mcpServers": {
    "varity": {
      "command": "npx",
      "args": ["-y", "@varity-labs/mcp"]
    }
  }
}

Note

Node.js 22.11 or later and Python 3.11 or later are required. If you do not have varitykit installed, your AI editor will prompt you to install it: pipx install varitykit.

Deploy with one prompt

Once the MCP server is installed, type this in your AI editor:

Deploy this project to Varity

Your AI editor calls varity_build and varity_deploy in sequence and returns the live URL. Varity builds from the project’s GitHub repository, so the project needs a GitHub origin. Without one, the deploy stops with NO_DEPLOY_SOURCE; ask your AI editor to push the project with varity_create_repo (this needs the GitHub CLI signed in or a GITHUB_TOKEN).

For a brand new project, ask your AI editor to build the app first, then deploy it:

Build a Next.js app called my-app with a dashboard, install dependencies, then deploy it to production

Your AI editor writes the app code, then the Varity tools handle dependency installation, the production build, and the deploy. You get a live URL when it is done.

Copy-paste prompts

These prompts are tested and work out of the box in Claude Code, Cursor, and Windsurf.

Check environment

Check if my environment is ready for Varity

Build a new app

Build a Next.js app called my-project with a dashboard, install dependencies, start the dev server, and open it in my browser

Deploy to production

Build my project and deploy it to Varity. Show me the live URL when done.

Check status

Show me the status of my Varity deployments

Debug a running app

Show me the runtime logs for my last Varity deployment

Estimate cost

How much does it cost to host this app on Varity?

Full build-to-deploy flow

Check my Varity environment, then build this project, deploy it, and open the live URL in my browser

What the MCP server can do

The Varity MCP server gives your AI editor a full set of tools. Here are the most useful ones for the build-to-deploy workflow:

Tool	What to ask your AI
varity_doctor	“Check if my environment is set up for Varity”
varity_login	“Log in to Varity”
varity_search_docs	“Search the Varity docs for Docker image deploys”
varity_install_deps	“Install dependencies for this project”
varity_dev_server	“Start the dev server and open it in my browser”
varity_open_browser	“Open the deployed app in my browser”
varity_build	“Build this project for production”
varity_create_repo	“Create a GitHub repo and push this project”
varity_deploy	“Deploy this to Varity”
varity_deploy_status	“Show me my deployment status”
varity_deploy_logs	“Show me the runtime logs” (dynamic apps only; static sites have no runtime logs)
varity_set_env	Does not accept values; it points you to a secret-safe way to set them
varity_redeploy	“Redeploy this app in place”
varity_delete_deployment	“Tear down this deployment”
varity_migrate	“Migrate my Vercel app to Varity”
varity_cost_calculator	“How much does the web-app profile cost?”
varity_list_templates	“What templates can Varity deploy?”
varity_template_info	“What does this template need?”

Full reference at MCP Server Spec.

Add Varity context to your project

For Claude Code, add this to your project’s CLAUDE.md:

## Deployment

This project deploys to Varity. Use the Varity MCP server tools:
- Deploy: ask "Deploy this project to Varity"
- Check status: ask "Show me my Varity deployment status"
- Logs: ask "Show me the runtime logs for my last deployment"

For Cursor, save this as .cursor/rules/varity.mdc:

When the user asks to deploy, use the Varity MCP varity_deploy tool.
When the user asks to check deployment status, use varity_deploy_status.
When the user asks to see logs, use varity_deploy_logs.

This tells your AI editor to route deploy-related requests to the Varity tools automatically.

Typical build-to-deploy session

Here is a complete session from idea to live app, using only your AI editor:

1. Check your setup

   Check if my environment is ready for Varity

   The AI runs varity_doctor and tells you if anything needs fixing.

2. Build the app

   Build a Next.js app called task-tracker. I want a dashboard where users can create and manage tasks.

   Your AI editor writes the project code: a Next.js app with a dashboard for creating and managing tasks.

3. Install dependencies

   Install dependencies and start the dev server

   The AI runs varity_install_deps and varity_dev_server, then opens localhost:3000.

4. Build the feature

   Work with your AI editor to build the task feature:

   Add a tasks data model with title (string), status (string), and dueDate (string) fields. Include a dashboard page with a list and a create form.

5. Deploy to production

   Build and deploy this to Varity

   The AI calls varity_build and varity_deploy, then reports the live URL.

6. Share the link

   Show me the live URL for my deployment

   Done. Send it to anyone.

Migrating from Vercel

If you already have a Vercel app, one prompt migrates it to Varity:

Migrate my Vercel app at https://github.com/your-org/your-app to Varity

The AI calls varity_migrate, which clones the repository, removes Vercel-specific configuration, and deploys to Varity infrastructure. See Migrate from Vercel for details.

Browser-based AI tools (Claude.ai, ChatGPT)

The hosted endpoint mcp.varity.so is retired. Browser-based tools that only accept a remote MCP URL are not supported today; use the stdio package (npx -y @varity-labs/mcp) from Claude Code, Cursor, VS Code, or Windsurf, or the llms.txt context files for LLMs without MCP support.

Troubleshooting

“varity not found” in your AI editor

Re-add the server and restart:

# Claude Code
claude mcp add varity -- npx -y @varity-labs/mcp

Verify with claude mcp list.

Deploy fails with “varitykit not found”

pipx install varitykit

Then ask your AI to try again.

Cursor does not show the Varity tools

Check that .cursor/mcp.json is valid JSON, then restart Cursor. Cursor requires a full restart (not just reload) to pick up MCP config changes.

Server is registered but tools are not being called

Ask your AI directly:

Use the varity_doctor tool to check my environment

If the AI is not picking up the tools automatically, being explicit about the tool name always works.

See MCP Server Spec for the full troubleshooting section.

Next steps

• MCP Server Spec: The full tool reference
• Deploy a Next.js App: Full Next.js deploy walkthrough
• Migrate from Vercel: One-command Vercel migration
• Auto-wired Services: What gets provisioned automatically