ClawRouter: an agent-native LLM router using wallet signatures and USDC micropayments for autonomous AI operations

What it solves

ClawRouter is designed for autonomous AI agents that cannot perform human-centric tasks like signing up for accounts, entering credit cards, or managing API keys. It provides a way for agents to access a wide variety of LLMs and other AI services (image/video generation, voice calls) independently and autonomously using wallet signatures for authentication and USDC micropayments for billing.

How it works

ClawRouter runs as a local proxy (on port 8402) that acts as an OpenAI-compatible gateway. It uses a weighted scorer across 15 dimensions to route requests to the cheapest capable model locally in under 1ms, reducing API costs by up to 92%. Authentication is handled via locally generated wallet signatures rather than API keys, and payments are processed per-request using USDC via the x402 protocol on Base and Solana networks.

Who it’s for

It is primarily built for autonomous AI agents and developers building agentic workflows. It integrates as a plugin for the OpenClaw coding agent, but also works as a standalone proxy for clients like continue.dev, Cursor, and any OpenAI-compatible SDK.

Highlights

• Agent-Native Auth & Payment: Uses wallet signatures instead of API keys and USDC micropayments instead of credit cards.
• Smart Local Routing: Automatically selects the most cost-effective model from 55+ options based on 15-dimension scoring.
• Zero-Setup Free Tier: Offers 8 free NVIDIA models available immediately without signup or crypto balance.
• Multimodal Capabilities: Supports image generation, image editing (img2img), and AI video generation (including character consistency via RealFace).
• Integrated Tooling: Provides built-in capabilities for AI-powered outbound voice calls (via Bland.ai/Twilio) and a unified crypto data API (Surf) for on-chain SQL queries.