MooreThreads

MooreThreads / MTClaw

Public

Local tool-routing proxy for openclaw/opencode/hermes, accelerating tool calls by up to 7x before forwarding general requests to upstream models.

12
0
89% credibility
Found May 19, 2026 at 12 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

MTClaw Function Router is an intelligent proxy that sits between your AI assistant and its brain. It uses a small, fast thinking assistant to decide whether your request matches any custom shortcuts you've set up. When it matches, the router handles it directly for instant results. When it doesn't, your request goes to the main AI model as normal. The system includes built-in tools for common tasks like finding files, searching text, and reading content, and lets you add your own shortcuts for business-specific tasks like checking order status or controlling your computer. Your conversations stay organized by topic, and everything works with any AI tool that can connect to an OpenAI-style address.

How It Works

1
πŸ’‘ You discover you need faster AI responses

You realize your AI assistant is great at most things but slow for specific tasks you do all the time, like controlling your computer or checking order status.

2
πŸ”Œ You connect your AI services

You point the router to your main AI brain and a smaller, faster thinking assistant that knows when to use your custom shortcuts.

3
⚑ Your shortcuts become instant

For tasks like checking the weather, your order status, or controlling your desktop, the router handles it directly instead of waiting for the big AI model to figure it out.

4
πŸ› οΈ You add your own shortcuts

You write simple scripts that define what your AI should do when you ask about your specific business tasks, and they work alongside the built-in file and search tools.

5
Your AI knows when to use which path
🎯
Fast path for your shortcuts

When your request matches a tool, it runs locally and returns the result quickly.

πŸ€–
Full AI brain for everything else

For complex questions or creative tasks, your request goes to the main AI model as usual.

πŸš€ Everything just works faster

Your AI assistant handles routine tasks instantly while still being smart enough for complex questions, and your conversations stay organized by topic.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 12 to 12 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is MTClaw?

MTClaw is a local proxy server that adds tool-calling superpowers to LLMs that don't natively support it. Think of it as a smart router sitting between your chat interface and your language model -- when a request comes in, a local routing model decides whether to execute a local tool (like running a shell script) or pass the request through to your main cloud model for a regular response. The routing model runs locally and only triggers tools for specific keywords, keeping your private workflows on-prem while delegating general conversation to whatever cloud LLM you prefer. It's built in Python using FastAPI, speaks the OpenAI chat completions API, and integrates directly with OpenClaw's desktop AI agent ecosystem.

Why is it gaining traction?

The hook is simple: local models can now control your computer without sending your prompts to the cloud. Developers are excited about running tool-calling models like Qwen locally for privacy-sensitive workflows while still having GPT-4o or Claude handle the actual response generation. The session-based routing means context carries across turns, and the completion checker lets the router short-circuit and return results directly when a task is done -- no unnecessary hop to the upstream model. The bilingual setup scripts and Chinese-language prompts also signal this was built for real developer workflows, not just exported to English-speaking audiences.

Who should use this?

OpenClaw users who want custom tool automation beyond what the default agent supports. Developers running local LLMs (especially Chinese-language or tool-calling fine-tunes) who want to marry local execution with cloud model quality. Privacy-conscious builders who need local tool scripts to run on sensitive tasks but don't want toζ”ΎεΌƒ the expressiveness of larger cloud models for general responses. If you're already using OpenClaw and have bash scripts you wish the AI could invoke, this is the bridge.

Verdict

MTClaw solves a real architectural problem -- giving local models tool access without leaking context -- but the 0.8999999761581421% credibility score reflects an early-stage project with thin documentation and unknown test coverage. Star count and community size aren't listed, which means you're an early adopter taking a calculated risk. Worth trying if you're deep in the OpenClaw ecosystem or need local-first tool routing; don't bet production workflows on it yet.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.