Nat Knight

#til #llms #claude #modelcontextprotocol

I was checking out the Model Context Protocol, a spec from Anthropic that lets you expose external programs to an LLM, and discovered that you can extend the Claude desktop app in this way. There are a bunch of enterprisey tools (and you can write your own, more interesting stuff), but out-of-the box you can enable URL-fetching, which lets you grab and process arbitrary information with Claude.

To enable the plugin,

1. Install the Claude desktop app.
2. Enable MCP under “Settings” > “Developer” > “Edit Config”
3. Add details about the MCP plugins you want in the indicated file.

For fetch, you'd add:

{
  "mcpServers": {
    "fetch": {
      "command": "uvx",
        "args": [
          "mcp-server-fetch"
        ]
    }
  }
}

Anthropic has more docs on using MCP plugins from the desktop app, and a list of built-in plugins that you can set up with minimal effort.

I think there are some exciting possibilities here! I don't think it should be too hard to give the model access to reference material, searchable history—basically retrieval augmented generation but for small, bespoke uses that wouldn't warrant the effort of custom-building an RAG system.

Of course, you're sending all of that off to Anthropic, so it would be particularly interesting to find a system that can do this with a self-hosted LLM, where the privacy and power concerns aren't as severe.