---
id: arcade-mcp-server
version: "1.25.2"
license: MIT
license_treatment: permissive
maintenance: active
---

# arcade-mcp-server — Model Context Protocol (MCP) server framework for Arcade.dev

License: permissive · Maintenance: active · Downloads: 696.0K/mo

## What it is and what it does

Arcade MCP Server is a framework for building Model Context Protocol (MCP) servers that expose your tools to AI assistants like Claude and development tools like Cursor. It uses a FastAPI-like decorator-based API to define tools, then automatically handles the MCP protocol negotiation and transport (stdio for local clients, HTTP/SSE for remote ones). The package includes built-in support for tool discovery, environment-based configuration, and hot reload during development.

You install it, define your tools with the @app.tool decorator, and run the server—it handles the rest. It's designed for both development (with automatic reloading) and production deployment (Docker, systemd, PM2, cloud platforms). The framework collects anonymous usage data by default (opt-out via ARCADE_USAGE_TRACKING=0) to track server starts and reliability.

Use it for:

- Expose custom Python functions as tools available to Claude Desktop or other MCP-compatible AI assistants.
- Build a local tool server for Cursor or other development environments that support the MCP protocol.
- Deploy a production MCP server over HTTP/SSE to serve multiple remote AI clients simultaneously.
- Integrate existing Arcade packages as discoverable tools without manual registration.
- Develop and test MCP tools locally with hot reload, then deploy the same code to production.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Arcade MCP Server provides a FastAPI-like framework for building Model Context Protocol servers that expose tools to AI assistants and development tools through stdio and HTTP/SSE transports.

Yes. The package is actively maintained, production-stable, has low install friction, and solves a clear problem—exposing Python tools to MCP-compatible AI assistants. MIT licensing is permissive. No known vulnerabilities. Install if you need to build an MCP server or integrate tools with Claude, Cursor, or similar platforms.

## Install

pip install arcade-mcp-server
uv add arcade-mcp-server
poetry add arcade-mcp-server

## Installing arcade-mcp-server

Before you install: Low install friction with a pure-Python wheel. Active maintenance (released 4 days ago) and marked production-stable. Supports current Python versions (3.10–3.13).

License in practice: MIT license is permissive; you can use, modify, and distribute this package with minimal restrictions, provided you include the license notice.

Quickstart:

pip install arcade-mcp-server

from arcade_mcp_server import MCPApp
from typing import Annotated

app = MCPApp(name="my-tools", version="1.0.0")

@app.tool
def greet(name: Annotated[str, "Name to greet"]) -> str:
    return f"Hello, {name}!"

if __name__ == "__main__":
    app.run()

Verify before relying:

- Whether tool discovery automatically indexes all Arcade packages in the environment or requires explicit registration.
- Performance characteristics and concurrency limits under multi-client load.
- Specifics of the hot-reload mechanism and its interaction with long-running tools.

## Package facts

- License: MIT (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 696.0K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

model context protocol server, mcp server framework, ai assistant tool integration, claude desktop mcp, fastapi mcp server, tool discovery protocol, ai tool exposure framework, mcp-protocol, ai-integration, tool-server

[View on SkillFed](https://skillfed.io/packages/arcade-mcp-server) · [View on PyPI](https://pypi.org/project/arcade-mcp-server/)