MCP Agent

What is MCP Agent?

MCP Agent is a Launchbar action that integrates Langchain with the Model Context Protocol (MCP) to automatically complete tasks based on user prompts. It runs on macOS via the Launchbar launcher.

How to use MCP Agent?

Add a .env file with your LLM API key and create an mcp_config.json file to configure one or more MCP servers (example provided for filesystem and fetch servers). Typing "@" triggers autocomplete suggestions from suggestions.js, which dynamically loads only the MCP server associated with the tool name you specify.

Key features of MCP Agent

• Integrates Langchain and MCP for task automation
• Works as a Launchbar action on macOS
• Configurable MCP servers via mcp_config.json
• Autocomplete suggestions for MCP server names (triggered by “@”)
• Dynamically loads only the MCP server linked to the chosen tool

Use cases of MCP Agent

• Automate file system operations (e.g., read, write, list files) using the filesystem server
• Fetch and process web content via the fetch server
• Combine multiple MCP servers to create custom automated workflows

FAQ from MCP Agent

What does MCP Agent do?

It automatically completes tasks based on your prompt by using Langchain and MCP servers.

How do I configure MCP Agent?

You need a .env file with your LLM API key and an mcp_config.json file that defines your MCP servers and their arguments.

Which MCP servers are supported?

The README demonstrates a filesystem server (run via npx) and a fetch server (run via uvx). You can add any MCP server by following the same JSON format.

Can I customize the MCP server names?

Yes. The suggestions.js file has a keyword section you can modify to match the installed MCP servers or to change their display names.