CodeOtter
Built with CodeOtter, reviewed with CodeOtter. We use it on the project itself as we keep improving it.
Code review for your
team and AI agents.
Powered by Dual-Engine rubrics.
CodeOtter brings calibrated System One 0–100 gauges, binary merge gates, blast radius scoring, and deep LLM walkthroughs into one platform. Zero per-seat tax, and the freedom to run 100% air-gapped with local GGUF + laya sidecars.
One-stop platform for all your PR review & release gate needs.
Move from noisy PR nitpicks to calibrated engineering scores across System One rubrics, blast radius, binary merge gates, LLM walkthroughs, and local GGUF sidecars.
Explore DocsSystem One Scores. 6 calibrated 0–100 gauges (Quality, Blast Radius, Risk, Tests, Readability, PR Hygiene) per PR.
Merge Gates. 9 pre-merge safety gates (S1_GATES): Title, Description, Security, Complexity, Tests, Docs, Scope, Guidelines, and Issue Requirements.
BYOK & Hosted Engines. Bring your own key for Claude, OpenAI, MiniMax, or OpenRouter. Or run fully offline with local GGUF (DeepSeek, Qwen, Llama, Microsoft CodeReviewer).
Zero telemetry by construction. ELv2 licensed.
100% offline-ready. Local GGUF only.
Mascot-guided Dual-Engine review, enforcing AGENTS.md & CLAUDE.md automatically.
CodeOtter auto-detects AGENTS.md and CLAUDE.md at your repository root via the GitHub API, while Otter—your System One + LLM reviewer—scores every diff and flags real defects before merge.
Explore Dual-Engine & OtterWork with your agents
and more…
Fast Troubleshooting.
No Context Switching.
When a System One model is configured, it owns all 6 calibrated 0–100 scores and every one of the 9 pre-merge safety gates via POST /v1/systemone—the language model is never asked to guess numbers.
CodeOtter checks your repo root for AGENTS.md and CLAUDE.md via the GitHub API and enforces both automatically with a per-repo gear toggle.
Download llama-server (GGUF), ggmlc laya (System One), and Python CodeReviewer sidecars into PR_SCORER_DATA with automatic GPU-then-CPU fallback.
Every score, gate, verdict, walkthrough, and finding is normalized by judge() before PocketBase storage, with strict gate() RBAC on every endpoint.
Like CodeOtter? Sponsor us.
Your support helps us keep building and improving CodeOtter for everyone.
Sponsor CodeOtterOur own projects
We use what we build.
CodeOtter is part of how we build our own projects. Here is where we use it today.
BazaarBox
Our storefront platform. We use CodeOtter while building and maintaining BazaarBox.
VectorMotion
Our React dashboard blocks project. CodeOtter is part of our development workflow here, too.
Your questions,
answered.
Quick answers to the questions teams usually ask while evaluating codeotter.
We believe engineering teams should own their PR review pipeline without per-seat or per-repository fees. CodeOtter is free to self-host under the Elastic License v2 (ELv2). If you like the platform, sponsorship helps us keep building and improving it.
Email hello@dharmeshgurnani.com to get in touch. We appreciate your support.
Any company or team, anywhere in the world! We welcome service companies, software agencies, cloud/VPS/GPU hosting providers, devtool vendors, and engineering organizations. Sponsors also get prominent logo placement across our website, documentation, and GitHub README.
Yes! CodeOtter is free to self-host under the Elastic License v2 (ELv2) using Docker, PocketBase, and local GGUF / laya sidecars or BYOK API keys. Sponsorship is how organizations contribute to continued development while getting direct maintainer access and custom engineering work.