How does Open Code Review decide which files to skip?

Open Code Review uses an allowlist to decide what reaches the model. Version 1.12.3 adds two rules to it: secret paths are excluded from review, and per-environment `.env` files are protected. Before this release those files could be bundled into a review like any other changed file, which put their contents in front of the configured model provider.

Which model providers does Open Code Review work with?

Open Code Review is OpenAI- and Anthropic-compatible and also accepts a custom provider, so it can point at a self-hosted or gateway endpoint. Provider and model are set through an interactive setup with `ocr config provider` and `ocr config model`, which means the same review pipeline can be repointed at a different model without changing the rules.

Can Open Code Review run in CI instead of locally?

Yes. Open Code Review ships CI/CD integration alongside its local modes, and it runs in several shapes: a workspace review, a branch range, a single commit, or a full-file scan with no diff history at all. It also integrates with coding agents including Claude Code, Cursor and Codex, and supports resumable review sessions.

How is Open Code Review different from asking a coding agent to review a diff?

Open Code Review splits the job: deterministic engineering logic handles file selection, bundling and rule matching, while the LLM agent handles dynamic context retrieval. It carries a built-in multi-language ruleset covering NPE, thread-safety, XSS and SQL injection. Alibaba states it has served tens of thousands of developers and found millions of code defects over two years.

Alibaba · 2026-09-16 · major

Open Code Review v1.12.3 — secrets and .env files stay out of the review

Open Code Review v1.12.3 keeps secrets out of an AI review: the allowlist now excludes secret paths and protects per-environment .env files. A path-traversal bypass in the code_search and code_comment tools is also fixed.

Open Code Review repository card on GitHub

Alibaba's open-source code reviewer now refuses to send secret paths and per-environment .env files to the model.

Key specs

GitHub stars	28,892
Stars gained that day	2,756

Quick facts

Version	v1.12.3
Maker	Alibaba
Released	2026-09-16
License	Apache-2.0
Language	Go
Install	npm install -g @alibaba-group/open-code-review
Providers	OpenAI, Anthropic, custom

What is it?

Two allowlist features land in Open Code Review v1.12.3: secret paths are excluded from review, and per-environment `.env` files are protected. The tool itself is Alibaba's internal code reviewer, released under Apache-2.0, which reads a git diff and returns line-level comments through an OpenAI- or Anthropic-compatible model.

How does it work?

Open Code Review pairs deterministic pipelines with an LLM agent: engineering logic picks and bundles the files, and the agent handles context retrieval through file-reading and code-search tools. v1.12.3 hardens those tools — `code_search` and `code_comment` now normalize paths, which closes a traversal bypass — and recovers a tool-call JSON response that arrives with trailing content instead of failing the run.

Why does it matter?

A code reviewer reads the whole repository, so anything it can open it can also send to a model provider. Excluding secret paths and `.env` files at the allowlist level means an API key sitting in a developer's local environment file stops being part of the review payload. The project is at 28,892 stars and gained about 2,756 of them on the day v1.12.3 shipped.

Who is it for?

teams running AI code review in CI

Frequently asked questions

How does Open Code Review decide which files to skip?: Open Code Review uses an allowlist to decide what reaches the model. Version 1.12.3 adds two rules to it: secret paths are excluded from review, and per-environment `.env` files are protected. Before this release those files could be bundled into a review like any other changed file, which put their contents in front of the configured model provider.
Which model providers does Open Code Review work with?: Open Code Review is OpenAI- and Anthropic-compatible and also accepts a custom provider, so it can point at a self-hosted or gateway endpoint. Provider and model are set through an interactive setup with `ocr config provider` and `ocr config model`, which means the same review pipeline can be repointed at a different model without changing the rules.
Can Open Code Review run in CI instead of locally?: Yes. Open Code Review ships CI/CD integration alongside its local modes, and it runs in several shapes: a workspace review, a branch range, a single commit, or a full-file scan with no diff history at all. It also integrates with coding agents including Claude Code, Cursor and Codex, and supports resumable review sessions.
How is Open Code Review different from asking a coding agent to review a diff?: Open Code Review splits the job: deterministic engineering logic handles file selection, bundling and rule matching, while the LLM agent handles dynamic context retrieval. It carries a built-in multi-language ruleset covering NPE, thread-safety, XSS and SQL injection. Alibaba states it has served tens of thousands of developers and found millions of code defects over two years.

Try it

npm install -g @alibaba-group/open-code-review