Co-Review

About Co-Review

Co-Review is a free, open-source code review tool built on Eclipse Theia. It opens an entire repository as a review session, letting you comment on lines, files, or folders and receive agent responses drawn directly from the codebase. The tool also handles diff reviews, PR reviews, and design planning with Mermaid diagrams, and it runs locally with no accounts or telemetry.

Review

Co-Review addresses a specific friction point that emerges when working extensively with coding agents: the gradual loss of familiarity with your own codebase. Rather than reviewing changes diff by diff, you open the whole repository and interrogate it. The agent answers from the actual code, not from a chat history, and the review thread becomes a persistent map of the project.

Key Features

• Whole-repository audit - The agent writes a guide covering the codebase map, vocabulary, flows, and weak spots, with findings grouped by area. Coverage tracking shows which files you have read, and a file that changes after you mark it read drops back to unviewed.
• PR review with agent first pass - For someone else's pull request, the agent takes an initial pass. You decide which comments go back to the author, and Co-Review posts them to GitHub as a real review with line comments.
• Design-before-code planning - The tool generates L1, L2, and L3 design plans using Mermaid diagrams. You revise the plan through comments before any code is built.
• Commit-aware threading - PR reviews check out a snapshot in a separate worktree, so the agent reads from that pinned state. Repo audit guides are pinned to a commit SHA, while threaded answers read the code as it exists when you ask. Comments stay anchored to their lines through subsequent edits.
• IDE foundation without the clutter - Built on Eclipse Theia, Co-Review keeps navigation, search, Git, and language server support while stripping out everything unrelated to reviewing.

Pricing and Value

Co-Review is free and MIT licensed. It runs locally on your machine with no account required and no telemetry. Code is sent only to the agent you configure. The tool is distributed as a native package and installs via a shell script: curl -fsSL https://sdsvn.github.io/co-revie... | bash.

Pros

• Runs entirely on your machine with no account, no telemetry, and an MIT license.
• Review guides are pinned to a specific commit, and coverage tracking flags files that have changed since you last read them.
• Posts PR reviews directly to GitHub with line comments, after you filter what the agent produced.
• Built on Eclipse Theia, which means real IDE navigation, search, and Git support without unrelated interface elements.
• Works with any MCP-compatible agent, including Claude Code, Codex, Cursor, Gemini CLI, Zed, OpenCode, and Goose.

Cons

• The tool is not well suited for teams that require centralized review administration, role-based access controls, or audit logging across multiple reviewers.
• Repository audit guides live inside Co-Review's workspace directory rather than as a versioned file in the repo, which means the guide won't travel with the code unless you copy it manually.
• On very large repositories, the depth and usefulness of the generated guide may vary - the maker has not published benchmarks for repo size thresholds where the audit starts to lose resolution.

Co-Review fits solo developers and small teams who use coding agents heavily and want to regain a working mental model of their codebase. It's also practical for conducting structured design reviews before implementation, or for taking a first pass at incoming PRs when you want the agent to surface issues before you spend time on the diff. The local-first, no-telemetry design makes it straightforward to try without committing to a platform.