bionic-mcp

You run models in LM Studio. bionic-mcp turns them into tools Claude Code (or any MCP client) can call, so you can hand the cheap and bulky work to your own machine. Free, local, private, nothing leaves your Mac.

You, in Claude Code:  "classify these 50 log lines by severity"
                       ↓
             runs on your hardware, 0 API tokens

Install

git clone https://github.com/turner-moore/bionic-mcp.git && cd bionic-mcp && npm install && \
claude mcp add -s user bionic -- node "$(pwd)/index.js"

Then start LM Studio's local server (that's where the models are):

~/.lmstudio/bin/lms server start

{
  "mcpServers": {
    "bionic": {
      "command": "node",
      "args": ["/absolute/path/to/bionic-mcp/index.js"],
      "env": { "BIONIC_BASE_URL": "http://127.0.0.1:1234" }
    }
  }
}

Related MCP server: Local LLM MCP Server

Why not just point your client at the LM Studio API

Two things would eventually bite you. This server fixes both:

Tools

local.js is a second, optional server. It tries a faster on-device backend first (caix, Apple Core AI on the Neural Engine) and falls back to LM Studio when that's down.

claude mcp add -s user local -- node "$(pwd)/local.js"

Requirements

• Node 18+ (for built-in fetch)

• LM Studio with its local server running

• The capability card lives in models.json, edit it to match your own installed models.