houtini-lm

Overview

Houtini LM is a specialized MCP server designed to save tokens by delegating bounded tasks from Claude Code to local or cloud LLMs. It works with various endpoints including LM Studio, Ollama, vLLM, DeepSeek, Groq, Cerebras, and OpenRouter. The server intelligently routes tasks to appropriate models, tracks performance metrics, and provides model discovery capabilities. It focuses on offloading well-defined tasks like code review, test generation, and documentation while keeping complex reasoning tasks with Claude.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• chatmessage: string, system?: string, temperature?: number, max_tokens?: number, json_schema?: object, model?: string

  Send a task to an LLM and get an answer. Offloads bounded tasks like generating boilerplate or explanations.

  Try:Generate unit tests for this functionTry:Explain this Python code in plain EnglishTry:Draft a commit message for these changes
• custom_promptinstruction: string, system?: string, context?: string, temperature?: number, max_tokens?: number, json_schema?: object, model?: string

  Send a three-part prompt (system, context, instruction) to an LLM for consistent high-quality output.

  Try:Review this code for security issuesTry:Summarize this research paperTry:Convert this API spec to OpenAPI format
• code_taskcode: string, task: string, language?: string, max_tokens?: number, model?: string

  Built-in code analysis with pre-configured system prompts for finding bugs, explaining code, or writing tests.

  Try:Find bugs in this TypeScript functionTry:Explain this Python classTry:Generate unit tests for this Rust code
• code_task_filestask: string, files: string[], max_tokens?: number, model?: string

  Review multiple related files or large files directly from disk without context window limits.

  Try:Review these React components for consistencyTry:Analyze this entire codebase for architectural issuesTry:Refactor these helper functions

Comparable tools

Installation

# For Claude Code
claude mcp add houtini-lm -- npx -y @houtini/lm

# For Claude Desktop
Add to claude_desktop_config.json:
{
  "mcpServers": {
    "houtini-lm": {
      "command": "npx",
      "args": ["-y", "@houtini/lm"],
      "env": {
        "HOUTINI_LM_ENDPOINT_URL": "http://localhost:1234"
      }
    }
  }
}

FAQ

What types of tasks get offloaded to the local model?

Bounded, well-defined tasks like generating test stubs, code review, converting formats, generating mock data, and writing type definitions. Complex reasoning, tool access, and multi-file changes remain with Claude.

How does model routing work?

Houtini-LM queries your LLM server for available models, looks up their metadata on HuggingFace, and maintains a local database with model capabilities. It then scores loaded models against task types to automatically pick the best one for each task.

Compare houtini-lm with