Connect an AI agent with MCP

Open MCP setup in your Appfly account. Use the server address shown there. It points to /mcp on your Appfly instance.

The server uses Streamable HTTP. The signed-in version of Appfly supports OAuth and API-key authentication.

Add the server address in a compatible client, then choose its connection or authentication action. Appfly opens a sign-in and authorization page. Check the requested access before allowing the connection.

Use an API key

Create a key in Developer API. Copy it when it is shown and configure your client to send:

Authorization: Bearer YOUR_API_KEY

A key belongs to your account and can access all of your apps and spend credits. Keep it out of Git, public prompts, and shared screenshots. See authentication.

Claude Code

Use the address from your account setup:

claude mcp add --transport http appfly https://fly.appfly.dev/mcp

Then open /mcp in Claude Code and authenticate. For another Appfly instance, replace the URL with the address shown in that instance’s setup.

Cursor

Add a remote server in Cursor’s MCP settings or its .cursor/mcp.json file. For API-key authentication:

{
  "mcpServers": {
    "appfly": {
      "url": "https://fly.appfly.dev/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

For OAuth, omit the headers and use the client’s sign-in flow where supported.

Other clients

Use a client that supports remote MCP over Streamable HTTP and the authentication method you choose. A web-hosted client needs an address it can reach; it cannot reach your local localhost server.

After connecting, ask: “Check my Appfly plan and list my apps.” Then try a workflow.