MCPWorks lets you create serverless Python and TypeScript functions that AI assistants can discover and execute via the MCP protocol. This guide gets you running in 5 minutes.

What You'll Do

Register an account, connect your AI assistant (Claude Code, Codex, GitHub Copilot), and create + execute your first serverless function — all through natural language.

Step 1: Create Your Account (1 min)

Go to api.mcpworks.io/register and sign up with your email. You'll get 1,000 free executions/month — no credit card required.

Step 2: Copy Your .mcp.json Config (30 sec)

After login, the dashboard shows your namespace and API key. Copy the generated config:

{
  "mcpServers": {
    "myns-create": {
      "type": "http",
      "url": "https://myns.create.mcpworks.io/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    },
    "myns-run": {
      "type": "http",
      "url": "https://myns.run.mcpworks.io/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Paste this into your project's .mcp.json file (or ~/.claude/settings.json for global access).

Step 3: Create a Service (30 sec)

Ask your AI assistant:

"Create a service called 'utils' in my MCPWorks namespace"

The AI will call make_service automatically.

Step 4: Create Your First Function (1 min)

Ask your AI assistant:

"Create a hello-world function in my utils service using the hello-world template"

Or create something custom:

"Create a function called 'word-count' that takes text and returns the word count"

The AI will write the code and call make_function with backend code_sandbox.

Step 5: Execute It (30 sec)

Ask your AI assistant:

"Execute hello-world with name 'MCPWorks'"

The function runs in a secure sandbox. You'll see the result immediately.

Available Templates

MCPWorks includes 5 starter templates you can use with make_function:

• hello-world — Greet by name, prove the system works
• csv-analyzer — Parse CSV data and return summary statistics
• api-connector — Call external APIs and return responses (requires httpx)
• slack-notifier — Post messages to Slack webhooks (requires httpx)
• scheduled-report — Generate markdown or JSON reports

Use list_templates to see all options, or describe_template for full details including code and schemas.

What's Happening Behind the Scenes

• Create endpoint (*.create.mcpworks.io) — manages your namespaces, services, and functions. Not metered.
• Run endpoint (*.run.mcpworks.io) — executes functions in a secure nsjail sandbox. Each execution counts against your monthly quota.
• Each function runs in an isolated container with no network access to your database or secrets.
• Python: 59+ pre-installed packages (numpy, pandas, httpx, and more) — use list_packages to see all.
• TypeScript: Full Node.js 20+ runtime with npm package support.

Next Steps

• Browse templates: ask your AI to list_templates
• See available packages: ask your AI to list_packages
• Read the full guide: Platform Guide covers namespaces, services, code mode, versioning, and more
• LLM agent reference: LLM Agent Reference has structured tool documentation for AI agents