Wrangler MCP — Cloudflare Workers for AI Agents

What it is

Wrangler MCP is a Model Context Protocol server that exposes Cloudflare's developer platform to AI agents. It wraps the Wrangler CLI, enabling Claude Code and other MCP-compatible agents to deploy Workers, read and write KV namespaces, manage R2 object storage, and query D1 databases through tool calls.

This MCP server targets developers who use AI agents for infrastructure management. Instead of switching to a terminal to run wrangler commands, you tell your agent what you need and it handles the Cloudflare API interactions.

How it saves time or tokens

Managing Cloudflare resources through the dashboard or CLI requires context-switching from your coding environment. With Wrangler MCP, your AI agent handles deployments, KV writes, and D1 queries inline during a coding session. A single natural language instruction like 'deploy this worker to production' triggers the correct sequence of wrangler commands.

How to use

1. Install the Wrangler MCP server in your Claude Code configuration.
2. Authenticate with your Cloudflare API token.
3. Use natural language to manage Cloudflare resources.

// Claude Code MCP configuration
{
  "mcpServers": {
    "wrangler": {
      "command": "npx",
      "args": ["wrangler", "mcp", "server"],
      "env": {
        "CLOUDFLARE_API_TOKEN": "your-api-token"
      }
    }
  }
}

Example

// Example Worker deployed via Wrangler MCP
export default {
  async fetch(request: Request, env: Env): Promise<Response> {
    // Read from KV
    const value = await env.MY_KV.get('config');

    // Query D1 database
    const result = await env.DB.prepare(
      'SELECT * FROM users WHERE active = ?'
    ).bind(1).all();

    return Response.json({
      config: value,
      users: result.results
    });
  }
};

Related on TokRepo

• MCP integrations — Other MCP servers for infrastructure management
• DevOps tools — Deployment and infrastructure automation tools

Common pitfalls

• The Cloudflare API token must have the correct permissions for each resource type (Workers, KV, R2, D1). A token with only Workers permissions cannot manage KV namespaces.
• Wrangler MCP executes real deployments. There is no dry-run mode by default. Test against a staging Cloudflare account before using it on production resources.
• Large R2 object uploads through MCP may time out. For bulk data transfers, use the Wrangler CLI directly with streaming support.