houtini-ai/houtini-lm

View on GitHub ↗

MCP server that saves Claude Code tokens by delegating bounded tasks to local or cloud LLMs. Works with LM Studio, Ollama, vLLM, DeepSeek, Groq, Cerebras.

107 ★26 forksJavaScriptUpdated 20d ago

What you need to know

Bounded task offloading from Claude Code to a local LLM to save tokens

Install

npm

Usage

  • Configure local LLM and offload bounded tasks

Key features

  • token savings
  • local LLM task offload

Best for

Teams wanting to cut Claude Code token spend for bounded tasks

Caveats

  • Supports only bounded/self-contained tasks
Platforms: Node.jsClients: Claude Code

Reviewed 2026-08-11

Topics

ai-agentsclaudeclaude-mcpcode-generationdeveloper-tooldeveloper-toolslm-studiolm-studio-mcplocal-llmmcpmcp-clientmcp-servermodel-context-protocolollamaopenai-apitoken-savingsvibe-coding-assistantvllm
Stars
107★
Forks
26
Language
JavaScript
License
Apache-2.0
Created
2025-09-04
Last push
2026-08-15