Repolode
Submit
BlockRunAI

BlockRunAI/ClawRouter

An LLM router for autonomous agents that picks the cheapest capable model per request locally, paid via a crypto wallet in USDC or a credit-card API key.

Open-source alternative to:OpenRouterLiteLLM

View on GitHubWebsite
Stars
6,633
Forks
650
Open issues
2
Last commit
yesterday

What is it?

ClawRouter is an open-source LLM router that sits between an AI client or agent and many model providers. It scores each request across 15 dimensions, filters out models that cannot handle the task (for example, no tool calling or no vision), and ranks the rest by quality, cost, speed and reliability. Routing runs entirely on the local machine, and the project exposes a local OpenAI-compatible proxy on port 8402 that tools such as continue.dev, Cursor and the OpenAI SDK can point at.

The project targets agents that cannot create accounts or enter card details. A wallet generated on first run acts as the identity, and each paid call settles in USDC over the x402 protocol on Solana or Base. Human users can instead sign up at BlockRun, top up with a credit card and use a bearer API key. Seven open-weight models are available with no signup or wallet.

Beyond chat, the router covers image generation, image editing, video generation and outbound AI phone calls through slash commands and HTTP endpoints. It is MIT licensed and ships as the npm package @blockrun/clawrouter, with an OpenClaw plugin installer.

Key features

  • 01Scores each request across 15 dimensions and routes to a capable model locally.
  • 02Filters out models that lack required tools, vision or context length before ranking.
  • 03Runs a local OpenAI-compatible proxy on port 8402.
  • 04Accepts USDC micropayments per request over x402 on Solana or Base.
  • 05Supports credit-card top-ups with an API key as an alternative to a wallet.
  • 06Includes seven free open-weight models that need no signup.

Use cases

  • Letting autonomous agents call LLMs without human-managed accounts or API keys.
  • Routing coding assistants like Cursor or continue.dev to cheaper models automatically.
  • Keeping costs low by sending simple requests to budget models and complex ones to stronger models.
  • Generating images or videos and placing AI voice calls from an agent workflow.

Quick start

  1. 1

    Install ClawRouter from npm (package @blockrun/clawrouter) using the install method in the README

  2. 2

    Run ClawRouter and pin a free model such as free/nemotron-3.5-lightning, no signup or wallet needed

Things to know

  • Paid models require either USDC in a wallet or a credit-card-funded API key from BlockRun.
  • The desktop control plane app is documented for macOS only.

Topics

#llm-router#ai-agents#x402#usdc#openai-compatible#cost-optimization#typescript#ai#anthropic#deepseek

FAQ

What is ClawRouter?
An LLM router for autonomous agents that picks the cheapest capable model per request locally, paid via a crypto wallet in USDC or a credit-card API key. ClawRouter is an open-source LLM router that sits between an AI client or agent and many model providers. It scores each request across 15 dimensions, filters out models that cannot handle the task (for example, no tool calling or no vision), and ranks the rest by quality, cost, speed and reliability. Routing runs entirely on the local machine, and the project exposes a local OpenAI-compatible proxy on port 8402 that tools such as continue.dev, Cursor and the OpenAI SDK can point at.
Is ClawRouter free to use? What license does it have?
ClawRouter is open source under the MIT license. Check the license terms before using it in commercial products.
Is ClawRouter still maintained?
Actively maintained. The last commit to ClawRouter was on Oct 21, 2026.
What can I use ClawRouter for?
Letting autonomous agents call LLMs without human-managed accounts or API keys. Routing coding assistants like Cursor or continue.dev to cheaper models automatically. Keeping costs low by sending simple requests to budget models and complex ones to stronger models. Generating images or videos and placing AI voice calls from an agent workflow.
How do I install ClawRouter?
Install ClawRouter from npm (package @blockrun/clawrouter) using the install method in the README Run ClawRouter and pin a free model such as free/nemotron-3.5-lightning, no signup or wallet needed
Where can I get ClawRouter?
The source code and installation instructions are on GitHub: https://github.com/BlockRunAI/ClawRouter

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects