mcp-local-llm

mcp-local-llm provides a local, cost-optimized layer for delegating mechanical or bulk text tasks to a local large language model backend. Claude Code remains the decision-maker for what to delegate, while the local model handles high-volume work such as summarization, classification, extraction, and drafting. The server exposes tools that you can call from Claude Code (or any MCP-compatible client): local_summarize for bulk text summarization, local_draft for initial content generation, local_classify for tagging and sorting, local_extract for structured data extraction, local_transform for formatting and style changes, local_complete for raw completions, and local_status to verify connectivity and available models. This separation helps reduce API usage costs while keeping Claude in charge of quality control and decision-making.

Prerequisites:

1. Ollama installed and running (local LLM backend).
2. Node.js 18+ installed.
3. Claude Code (or any MCP-compatible client).

Installation steps:

1. Install Ollama and pull a model (example):
   • brew install ollama
   • ollama serve
   • ollama pull qwen2.5-coder:7b
2. Clone the repository and install dependencies:
   • git clone https://github.com/aplaceforallmystuff/mcp-local-llm.git
   • cd mcp-local-llm
   • npm install
3. Build the project:
   • npm run build
4. Run the MCP server (example):
   • node dist/index.js
5. Add the MCP server to Claude Code or your MCP client:
   • In Claude Code: claude mcp add local-llm -s user -- node /path/to/mcp-local-llm/dist/index.js
   • Or edit ~/.claude.json to include: { "mcpServers": { "local-llm": { "command": "node", "args": ["/path/to/mcp-local-llm/dist/index.js"] } } }
6. Verify connectivity:
   • Use the local_status tool in Claude Code to confirm Ollama connection and available models.

Environment variables are optional and default values work with a standard Ollama setup. If you prefer a Docker-based OpenAI-compatible backend, you can use the Docker Model Runner and point LOCAL_LLM_BASE_URL to the TCP endpoint. Common issues include ensuring Ollama is running (ollama list), matching the model name with LOCAL_LLM_MODEL, and adjusting LOCAL_LLM_MAX_TOKENS or LOCAL_LLM_TEMPERATURE to fit your workload. The Delegation Philosophy section explains what tasks are best suited for local delegation versus Claude's core capabilities.