3.2K ·GitHub
What is this?
Review is an open-source desktop app that turns a branch or pull request produced by a coding agent into a guided, interactive review, letting you explore the architecture, inspect diffs, and ask questions about the generated code.
Review is an open-source desktop app that turns a branch or pull request produced by a coding agent into a guided, interactive review, letting you explore the architecture, inspect diffs, and ask questions about the generated code.Is this a general code editor or a dedicated review tool?It is a purpose-built review app, not a general editor. It is built on a vendored fork of Code OSS (the open-source base of VS Code) stripped down for the single job of reviewing AI-generated diffs.
It is a purpose-built review app, not a general editor. It is built on a vendored fork of Code OSS (the open-source base of VS Code) stripped down for the single job of reviewing AI-generated diffs.How do I get it running?Download the Review app and open it, then connect Claude Code, Codex, or another coding agent from the welcome screen and ask the agent to review your current branch against main.
Download the Review app and open it, then connect Claude Code, Codex, or another coding agent from the welcome screen and ask the agent to review your current branch against main.Which coding agents does it work with, and what do I need to build it?Review connects to Claude Code, Codex, and other coding agents from its welcome screen; building it from source requires Node.js 24 and pnpm 11.
Review connects to Claude Code, Codex, and other coding agents from its welcome screen; building it from source requires Node.js 24 and pnpm 11.What doesn't work yet?Review is still in beta: it only maps architectural changes within one repo, doesn't properly handle stacked PRs, and sharing reviews between machines is currently painful.
Review is still in beta: it only maps architectural changes within one repo, doesn't properly handle stacked PRs, and sharing reviews between machines is currently painful.Who is Review actually for?It's for developers who use AI to generate code but still want to understand how it works and catch regressions; if you're happy reading every diff line by line, it's probably not for you.
It's for developers who use AI to generate code but still want to understand how it works and catch regressions; if you're happy reading every diff line by line, it's probably not for you.Does Review send my code anywhere?Review runs against local checkouts and only sends anonymous telemetry, which excludes your code, diffs, comments, questions, prompts, and model output.
Review runs against local checkouts and only sends anonymous telemetry, which excludes your code, diffs, comments, questions, prompts, and model output.How is Review different from tools like Greptile, Bugbot, or CodeRabbit?Those tools have another LLM review your AI-generated code for you. Review is complementary: it doesn't judge your code, it helps you understand the diff, trace, and architecture behind it.
Those tools have another LLM review your AI-generated code for you. Review is complementary: it doesn't judge your code, it helps you understand the diff, trace, and architecture behind it.How is Review different from Plannotator?Plannotator is an alternative to the terminal for reviewing plans and simple diffs. Review takes lifecycle inspiration from Plannotator but is built as a more opinionated framework specifically for reviewing AI-generated code changes.
Plannotator is an alternative to the terminal for reviewing plans and simple diffs. Review takes lifecycle inspiration from Plannotator but is built as a more opinionated framework specifically for reviewing AI-generated code changes.Why vendor Code OSS instead of maintaining patches on VS Code?Coding agents struggle with applying patches, and a large chunk of stock VS Code's codebase is Copilot-related code Review doesn't need, so the team vendors Code OSS directly and merges in upstream security and feature fixes as needed.
Coding agents struggle with applying patches, and a large chunk of stock VS Code's codebase is Copilot-related code Review doesn't need, so the team vendors Code OSS directly and merges in upstream security and feature fixes as needed.How will the team monetize a free, self-hostable app?They plan to eventually charge companies for a hosted Review product with extras like trajectory storage and review sharing, while keeping the core app free and self-hostable.
They plan to eventually charge companies for a hosted Review product with extras like trajectory storage and review sharing, while keeping the core app free and self-hostable.What are similar projects?plannotator, pi-gui, CanvasTTY, monocode, zuse, bb, codex-host, wicked-studio
plannotator, pi-gui, CanvasTTY, monocode, zuse, bb, codex-host, wicked-studioWhich topics is it in?diff, code-editor, code-review, monorepo, apple-silicon, electron, agents, developers, coding-agent, codex, privacy, claude-code
diff, code-editor, code-review, monorepo, apple-silicon, electron, agents, developers, coding-agent, codex, privacy, claude-code