arcade
The best way to create, deploy, and share MCP Servers
claude mcp add --transport stdio arcadeai-arcade-mcp python server.py
How to use
Arcade MCP Server provides a Python-based MCP server framework with sample tools and seamless integration with Arcade tooling. The repository includes a generated server scaffold (server.py) with example tools such as greet, whisper_secret, and get_posts_in_subreddit, all wired into an MCPApp. You can run the server directly using Python or through the Arcade/uv wrapper for transport options. Once running, the server exposes a local API and a docs interface at http://127.0.0.1:8000/docs when using HTTP transport. The included examples demonstrate how secrets and authentication can be injected at runtime, how to call external APIs securely, and how to structure tools with annotations and context objects for robust MCP tool development.
How to install
Prerequisites:
- Python 3.9+ (or a version supported by Arcade MCP Server)
- Git
- (Optional) make utility if you follow the repository's install script
Installation steps:
- Clone the repository: git clone https://github.com/ArcadeAI/arcade-mcp.git && cd arcade-mcp
- Install dependencies (via Makefile convenience):
make install
or install Python dependencies manually, e.g.:
python -m pip install -r requirements.txt
- Run the server:
Run directly with Python (_stdio transport by default)
python server.pyOr run with HTTP transport (docs available at /docs):
python server.py http - (Optional) Configure MCP clients to connect to the server as demonstrated in the README (arcade configure commands).
Additional notes
Tips and notes:
- The server ships with example tools illustrating how to handle secrets, authentication, and external API calls. Inspect server.py for the tool definitions and wiring.
- When running with HTTP transport, the built-in API docs are available at http://127.0.0.1:8000/docs.
- Secrets can be injected at runtime via environment variables or a .env file. Use context.get_secret(...) to retrieve them inside tools.
- If you prefer using the Arcade CLI (arcade tool), you can configure clients to connect to a stdio server in your current directory or point tools to a local HTTP endpoint depending on the transport you choose.
- The repository's install path (make install) may install additional dependencies or set up scripts; refer to the Makefile for details.
- This is a Python-based MCP server; if you want to switch to a different transport or deployment method, you can adjust the command and transport you pass to the MCP client.
Related MCP Servers
mcp-use
The fullstack MCP framework to develop MCP Apps for ChatGPT / Claude & MCP Servers for AI Agents.
agentgateway
Next Generation Agentic Proxy for AI Agents and MCP servers
agent
A unified Model Context Protocol server implementation that aggregates multiple MCP servers into one.
pluggedin-app
The Crossroads for AI Data Exchanges. A unified, self-hostable web interface for discovering, configuring, and managing Model Context Protocol (MCP) servers—bringing together AI tools, workspaces, prompts, and logs from multiple MCP sources (Claude, Cursor, etc.) under one roof.
CanvasMCPClient
Canvas MCP Client is an open-source, self-hostable dashboard application built around an infinite, zoomable, and pannable canvas. It provides a unified interface for interacting with multiple MCP (Model Context Protocol) servers through a flexible, widget-based system.
mcpcat-python-sdk
MCPcat is an analytics platform for MCP server owners 🐱.