MCP-SuperAssistant

MCP SuperAssistant provides a Chrome extension that integrates MCP tools with a wide range of AI platforms (ChatGPT, Google Gemini, Perplexity, Grok, Google AI Studio, OpenRouter, Kimi, GitHub Copilot, Mistral, Qwen, Z Chat, and more). To use it locally, run the MCP SuperAssistant proxy in conjunction with a local or remote MCP server configuration. The proxy connects your AI platform to MCP tools via a configured transport (default SSE in the example). Once the proxy is running and connected to a server, you can invoke MCP tools from supported platforms and have results returned into your chat interface. The extension itself handles tool detection, execution, and result insertion, while the proxy forwards tool calls to the appropriate MCP server and returns responses.

Typical workflow: write or load a config.json describing your MCP servers, start the proxy, then connect your AI platform’s MCP sidebar to http://localhost:3006/sse (or the transport endpoint you choose). From there you can select and run MCP tools directly within the chat interface, see results rendered in the conversation, and leverage features like auto-execute and result insertion as configured.

Prerequisites:

• Node.js (14.x or newer) and npm installed on your system
• Basic familiarity with running commands in a terminal

Step 1: Create a working directory for your MCP setup

• mkdir mcp-setup && cd mcp-setup

Step 2: Prepare a config.json describing your MCP servers

• Create a file named config.json with your MCP server definitions. Example: { "mcpServers": { "desktop-commander": { "command": "npx", "args": [ "-y", "@wonderwhy-er/desktop-commander" ] } } }

Step 3: Install and run the MCP proxy via npx

• This uses the MCP SuperAssistant proxy package
• Run one of the following commands (example using SSE): npx -y @srbhptl39/mcp-superassistant-proxy@latest --config ./config.json --outputTransport sse

Optional transports you can use instead:

• npx -y @srbhptl39/mcp-superassistant-proxy --config ./config.json --outputTransport streamableHttp
• npx -y @srbhptl39/mcp-superassistant-proxy --config ./config.json --outputTransport ws

View all available options:

• npx -y @srbhptl39/mcp-superassistant-proxy@latest --help

Step 4: Connect your AI platform

• Open the MCP SuperAssistant sidebar in your chosen platform and connect to the local proxy endpoint, typically http://localhost:3006/sse for SSE.
• Ensure the server status shows Connected before trying to run tools.

Tips and common issues:

• If you change config.json, restart the proxy to pick up changes.
• Use a local MCP server for best performance and fewer CORS issues; remote servers can be proxied via the same config.
• Ensure your transport endpoint matches what you configured (SSE, Streamable HTTP, or WebSocket).
• The npm package used here is @srbhptl39/mcp-superassistant-proxy. If you’re testing locally, you can run it with npx as shown in the steps above.
• For debugging, check the proxy’s health endpoints and logs to verify connectivity to the MCP servers.
• You can add multiple MCP servers under mcpServers to multiplex tool calls across sources.
• If your environment blocks the default ports, configure the proxy to listen on an alternate port as supported by the proxy tool.