Review with any model — including open-weight
CodeMouse reviews your GitHub pull requests with whichever AI models you choose. Bring your own API key, pay the provider directly with zero markup, and connect several models at once so they cross-check each other.
Supported providers
- Anthropic — Claude Opus, Sonnet and Haiku.
- OpenAI — GPT-5.3 Codex and GPT-5 Codex.
- Google — Gemini 2.5 Pro, Flash and Flash-Lite.
- OpenRouter — one key for open-weight models including Kimi, DeepSeek, GLM, Qwen and 300+ more.
Open-weight models via OpenRouter
Add a single OpenRouter key and pick any of these (or paste any other OpenRouter model id). Prices are $ per 1M tokens, input · output:
- Kimi K3 (moonshotai/kimi-k3) — latest Moonshot frontier model, top SWE-bench. $3.00 · $15.00
- Kimi K2.7 Code (moonshotai/kimi-k2.7-code) — code-specialized, excellent value. $0.73 · $3.50
- GLM-5.2 (z-ai/glm-5.2) — top open-weight intelligence, strong agentic coding. $0.76 · $2.39
- DeepSeek V4 Pro (deepseek/deepseek-v4-pro) — strong reasoning, MIT-licensed. $0.44 · $0.87
- Qwen3 Coder Next (qwen/qwen3-coder-next) — cheapest and fastest coder. $0.11 · $0.80
Many models, one consensus
Select two or three models and each reviews the PR independently. CodeMouse merges their findings, de-duplicates overlap, and only auto-approves when every model agrees — so more models means a stricter gate, not noisier comments. Inference is billed straight to your key; CodeMouse adds nothing on top.
Start a 14-day free trial — $10/month flat for the whole team, no card required.