DEV.co
AI Frameworks · BlockRunAI

ClawRouter

ClawRouter is an open-source LLM router that automatically selects and routes requests to the cheapest capable model from 55+ options (OpenAI, Anthropic, Google, DeepSeek, etc.) in under 1ms. It's designed for autonomous AI agents, using wallet signatures instead of API keys and USDC micropayments via the x402 protocol instead of credit cards. Eight models are free with no signup required.

Source: GitHub — github.com/BlockRunAI/ClawRouter
6.6k
GitHub stars
625
Forks
TypeScript
Primary language
MIT
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
RepositoryBlockRunAI/ClawRouter
OwnerBlockRunAI
Primary languageTypeScript
LicenseMIT — OSI-approved
Stars6.6k
Forks625
Open issues39
Latest releasev0.12.212 (2026-06-23)
Last updated2026-07-05
Sourcehttps://github.com/BlockRunAI/ClawRouter

What ClawRouter is

A TypeScript-based local routing proxy that analyzes requests across 15 dimensions, scores against 55+ LLM models, and forwards to the optimal provider. Runs on port 8402 as OpenAI-compatible endpoint. Supports wallet-based auth (no API keys), USDC settlement on Base and Solana via x402 protocol, and integrates with OpenClaw agents. Routing logic is open-source and executes locally with no external routing dependencies.

Quickstart

Get the ClawRouter source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/BlockRunAI/ClawRouter.gitcd ClawRouter# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Autonomous AI agents with cost constraints

Agents that need to operate independently without account signup or credit card payment. ClawRouter's wallet-signature auth and per-request USDC micropayments align with agent-native design patterns. Reduces per-request costs by selecting the cheapest capable model automatically.

Multi-model inference with cost optimization

Applications requiring access to 55+ models (GPT-4, Claude, Grok, Gemini, DeepSeek, etc.) with automatic cost-aware routing. Useful for workloads where model capability requirements vary by request type, allowing the router to select appropriately priced alternatives at runtime.

Development environments with zero signup friction

Eight free models (NVIDIA, Mistral, Qwen) work immediately out of the box with no account, API key, or payment method. Ideal for prototyping, CI/CD pipelines, and local development where API keys are undesirable or agent authentication is needed.

Implementation considerations

  • Installation has two paths: one-shot OpenClaw script (recommended, handles full plugin registration and wallet setup) or npm + manual setup. Skipping `clawrouter setup` leaves OpenClaw in a broken state with only 7 hardcoded models instead of ~38. Clear guidance exists but requires careful reading.
  • Wallet generation is automatic on first run; printing to stdout. For paid models, manually send USDC to that address on Base or Solana. No balance requirement for free tier, but cold-start experience is: install → run → pick a free model → optional: send USDC later.
  • Proxy runs locally on port 8402 with OpenAI-compatible `/v1/chat/completions` endpoint. apiBase must end with `/v1/` (trailing slash critical) for clients like continue.dev; omitting it causes 404 errors. Straightforward for tools with OpenAI provider support.
  • Free tier (8 NVIDIA/Mistral/Qwen models) is fully operational with zero friction. Paid models require x402 protocol support and wallet funding. Switching between free and paid tiers is transparent but requires understanding USDC micro-economics.
  • Routing is deterministic based on 15 request dimensions and model capability scores; no randomness or fallback logic is documented. Behavior if a selected model is unavailable or returns an error is not stated in the provided README.

When to avoid it — and what to weigh

  • Guaranteed low-latency routing (<1ms unacceptable) — While ClawRouter claims <1ms routing locally, any added network latency to the underlying model provider compounds request time. If sub-millisecond end-to-end latency is critical, the routing layer itself is negligible but provider selection speed depends on local compute and 15-dimension scoring.
  • Cryptocurrency or USDC payments are forbidden or unknown — Paid model access requires USDC on Base or Solana. Organizations with crypto-hostile compliance, procurement policies, or unfamiliarity with stablecoin wallets will face friction. Free tier (8 models) avoids this, but broader model access requires crypto setup.
  • Closed-source, audited routing logic is required — ClawRouter's routing algorithm is open-source but may not suit organizations requiring proprietary, vendor-audited, or compliance-certified routing policies. The 15-dimension scoring is visible but not formally certified for security or fairness properties.
  • Models not in the 55+ curated list are needed — ClawRouter supports a fixed set of 55+ models; no custom provider integration or BYOK (bring-your-own-keys) for unlisted models. If you require specific proprietary or niche models outside this list, you'll need a different router (e.g., LiteLLM, OpenRouter).

License & commercial use

MIT License (permissive, OSI-approved). Permits commercial use, modification, and redistribution under MIT terms. No trademark restrictions stated. No copyleft obligations.

MIT license permits commercial use without explicit restrictions. However, reliance on x402 protocol (not clearly defined in README), USDC settlement on external blockchains (Base, Solana), and integration with OpenClaw (third-party framework) introduces dependencies outside ClawRouter's direct control. Verify x402 protocol terms, Base/Solana network compatibility, and OpenClaw licensing for your deployment. No enterprise SLA, support, or indemnification is stated. Requires review of complete license file and x402 protocol terms before commercial deployment.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityClear
Deployment complexityModerate
DEV.co fitGood
Assessment confidenceMedium
Security considerations

Wallet-based authentication (signature instead of API keys) reduces exposure of long-lived secrets; signatures are ephemeral. Local routing logic (no external dependencies for routing) reduces attack surface for the routing decision itself. However, wallet private key management is not discussed—users must secure local wallet files. USDC micropayments over public blockchains (Base, Solana) are transparent but immutable; transaction mistakes cannot be reversed. x402 protocol details are not provided in README; review x402 spec for payment validation, dispute resolution, and fraud prevention. No mention of rate limiting, DDoS protection, or input validation. Depends on OpenAI, Anthropic, Google, DeepSeek, etc. upstream; their security posture is outside ClawRouter's control. Suggest security audit before production use, especially for agents with financial autonomy.

Alternatives to consider

OpenRouter

200+ models, mature platform, credit card payment (no crypto required). Better for teams avoiding blockchain infrastructure or needing a hosted, managed service. Manual model selection (not smart routing).

LiteLLM

Open-source, 100+ models, runs locally. Bring-your-own-keys model (you manage multiple API keys). Better if you already have accounts with multiple providers and want local routing without crypto. Less agent-native (requires API keys, not wallet signatures).

Portkey

Hosted gateway, observability-first, supports 100+ models. $49–499/mo pricing. Better for enterprises needing managed infrastructure, SLAs, and fallback/retry policies. Crypto-agnostic; no blockchain dependency.

Software development agency

Build on ClawRouter with DEV.co software developers

Install the one-shot OpenClaw plugin or run as a standalone proxy. Free tier (8 models) works immediately. Add USDC later for paid models. No signup, no API keys, no credit card required.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

ClawRouter FAQ

Do I need crypto or USDC to use ClawRouter?
No for free tier. Eight models (NVIDIA, Mistral, Qwen) work immediately with zero setup, no crypto, no API key. For paid models, you send USDC to your auto-generated wallet address on Base or Solana. Free tier is fully functional for prototyping and development.
What happens if I skip `clawrouter setup` with OpenClaw?
OpenClaw will only show ~7 hardcoded models instead of the full ~38 BlockRun models. The plugin won't be registered, models allowlist won't sync, and auth profile won't be written. Run `clawrouter setup` to repair, or use the one-shot install script to avoid this.
How fast is the routing?
ClawRouter claims <1ms local routing (15-dimension scoring and model selection). This is the time to choose a model; actual inference latency depends on the upstream provider (OpenAI, Anthropic, etc.) and network conditions.
Can I use ClawRouter without OpenClaw?
Yes. ClawRouter runs as a local proxy on port 8402 with an OpenAI-compatible `/v1/chat/completions` endpoint. Any client supporting OpenAI-compatible APIs (continue.dev, Cursor, VS Code) can point to `http://localhost:8402/v1/` with API key `x402`.

Software development & web development with DEV.co

From first prototype to production, DEV.co delivers software development services around tools like ClawRouter. Our software development agency staffs experienced software developers and web developers for custom software development, web development, integrations, and ongoing support across ai frameworks and beyond.

Get Started with ClawRouter for AI Agent Routing

Install the one-shot OpenClaw plugin or run as a standalone proxy. Free tier (8 models) works immediately. Add USDC later for paid models. No signup, no API keys, no credit card required.