@github.awesome: ClawRouter is a Smart LLM router for your OpenClaw agent that stops you from burning money. If it’s a basic question, it routes it to a cheap model like DeepSeek; if it’s complex code, it escalates to Claude Sonnet. It runs entirely locally, supports micropayments so you don't need to manage thirty different API keys, and finally cuts your inference bill by 80% without you lifting a finger. #github #opensource

Andre Costa :

clawdbot is just a wast of time. It is way faster to build the agentic flow you need and then just start

` :

Can it reroutes to my github copilot that I linked to openclaw instead of another one on the internet for complex tasks? Like using copilot for complex tasks and deepseek for simple ones

Justin Griffith :

Openrouter is better

That stupid dragon🐾 :

does anybody know how to use ollama with openclaw

Bloody Kheeng :

qwen ai is free why dont people use it.

Flip :

30$ in 5 Hours for setup, one web search, telegram implementation and half the setup process of WhatsApp - just because I mistakenly let it run opus for everything. I’m back at Kimi-2.5 now which does the job for most of my tasks perfectly.

Emmanuel | YourFavoriteTechBro :

How does it know if the question is basic or not. What’s the ranking system

Deveyes :

OR run a decent coding llm locally and don't pay ANY inference bill

Meaty :

Is this malicious?

Luis :

It took me a day to tune a single installation to optimize token consumption using a hybrid approach routing questions through a tiered approach including local models consumption (no bill cost other than VM resources). Adding more components will increase your installation complexity and add latency.

mordor 🦁 :

Doesn't cursor have this functionality by default using the auto function?

Kendall :

How much are they skimming off the top of the token usage?

Vincent :

Try running something like a local 20B model. How good is it?

MirinfitnessDailyInspired🛞 :

takes one prompt to make this.

United_States_of_Twilightzone :

I set this up with my own agent using open router

mdlc.ai :

R :

or just use openrouter/auto

ToddMichaelTV :

thank you brother

IkiruFlowers :

Lovely

southerncharmjason :

Why not a local LLM like ollama for even cheaper ?

User4908632995480 :

W

Ag :

what about security?

p.diezzz__ :

noiicee

𝖛𝖎𝖛𝖊𝖘 𝐆𝐄𝐑𝐀𝐒𝐈𝐌𝐎𝐒 :

Excellent idea, as AI models will become like electricity, at the end all will become smart enough, we'll select the cheapest. Will intergrate a similar feature to the upcoming Plexon AI.

Newfonsz :

Openrouter🥰