Approving AI-generated code changes by scrolling through a diff in your editor is clunky. Crit, a new open-source terminal tool written in Go, gives you a dedicated review interface for exactly this workflow.
The tool launched this week at v0.2.0 and works in two modes. For documents, it opens markdown files in a full-screen terminal UI with syntax highlighting and a comment sidebar. For code, it detects changed files in a git repo, shows them in a tabbed interface with green gutter markers on modified lines, and lets you leave inline comments across multiple files.
The key workflow: your AI coding agent generates changes, Crit opens a review session, you leave comments on specific lines, and the agent picks up those comments to make edits. It's the "human-in-the-loop" step that most AI coding setups handle poorly.
Claude Code Integration
Crit ships with direct Claude Code support. You can install it through the Claude Code Plugin Marketplace or set it up manually with crit setup-claude. Once configured, running /crit:review launches the TUI, and Claude reads your comments to implement feedback.
The tool also exposes a scriptable CLI with JSON output, so other AI agents or automation can feed comments in programmatically. There's tmux integration for split-pane setups where you want the review alongside your agent's terminal session.
Navigation uses familiar Vim-style keybindings: j/k to scroll, enter to comment, v for visual select mode on multi-line selections, and tab to switch between file tabs during code review.
Crit requires Go 1.21+ to build from source (go install github.com/kevindutra/crit/cmd/crit@latest) and tmux for the Claude Code integration. It's MIT-licensed and entirely self-contained - no external services or API keys needed.