Point Codex at BlockRun and choose the model per task
clawrouter-codex translates Codex's Responses API onto BlockRun chat completions, so Codex runs on any model in the catalog and pays per request in USDC, no OpenAI key.
How it works
Codex speaks OpenAI's Responses API. clawrouter-codex is a small local front-adapter that accepts that shape, translates it onto chat completions, routes it through ClawRouter and pays for the call over x402. Codex sees an OpenAI-compatible endpoint; BlockRun sees a normal paid request.
The point is choice per task: a cheap model for boilerplate, a frontier model for the hard refactor, switched by the model string rather than by a second subscription. The open-weight models are free, so the adapter works before the wallet holds anything.
Behind it is the BlockRun catalog: 78 chat models plus image, video, music and speech generation, search, market data and multi-chain RPC, priced per call. 6 open-weight models are free and need no wallet.
Setup
1. Bring the adapter up
npx @blockrun/clawrouter-codex up2. Run Codex with the profile it wrote
codex --profile clawrouterQuestions
- Is this an OpenAI key replacement?
- Yes for the model calls: the adapter pays per request from a local wallet, so Codex needs no OpenAI key to reach the models BlockRun serves.
- Which models work?
- Any chat model in the catalog — OpenAI's, Anthropic's, Google's, DeepSeek's and the open-weight ones — selected by name.