diff-tour turns AI code changes into a guided review
diff-tour organizes git diffs into a read-only walkthrough, helping developers understand what must be fixed before merging changes from Claude Code or Codex.
If you cannot make sense of code that an AI agent has just changed, install diff-tour. It makes Claude Code or Codex walk you through the changes the way an experienced developer would read them.
First it explains why the changes are needed, then it shows the files worth reading, including related files that were not modified.
It asks the questions for you: “Does what used to work still work?”, “What happens during a failure or rerun?”, “Can the changes be safely rolled back?”
It separates must-fix issues from optional improvements and explains why parts that look risky may actually be fine.
Where it makes sense to pause and think, it asks a question first so you can guess the answer before seeing the explanation.
Run it before merging, and you will understand where these changes could break something while learning how an experienced developer reads code.
The entire process runs in read-only mode. The Claude Code or Codex session you are already in prepares the review in two to five minutes. Every line of code shown comes directly from git.
The install script verifies the downloaded file with checksums and installs the tool into your user folder. On Mac or Linux, run the one-line command at the top of the README, then enter tour in your project.
When the review is done, pass the required fixes to Opus 5.5:
diff-tour reviewed my latest changes and flagged what must be fixed: [list them here]. For each item, explain how you plan to fix it and which files you will touch, then wait for my approval before changing anything. After the fixes, run the tests and report which items are fixed and which are not.
Why it matters
AnalysisIt reduces the risk of merging opaque AI-generated changes and teaches a structured code-review habit for agent-assisted development.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
The /stack mod for Claude Code: a full view of installed components
The /stack mod for Claude Code reveals all skills, subagents, commands, hooks, MCP servers, plugins, settings, and output styles grouped by configuration level.
ai-job-search for Claude Code: open-source AI tool for job applications
A GitHub project for Claude Code automates resumes, cover letters, vacancy matching and interview preparation with two AI agents.
Claude Code Projects: How a Single AI Orchestrates a Team of Autonomous Agents
A deep dive into Claude Code Projects: Coordinator and Threads architecture, parallel branches, shared memory via MEMORY.md, and cloud agent management.
MCP Servers and Claude Code: A Practical Guide to Connecting and Building
How to expand Claude Code's capabilities with Model Context Protocol: connecting the filesystem, GitHub, PostgreSQL, Brave Search, and developing a custom server in TypeScript.
Claude Code
The official terminal agent from Anthropic, operating directly in the command line via Claude 3.7 Sonnet with native support for Bash, Git, file systems, and the MCP protocol.
Patch & Diff-First Generation
A coding interaction pattern where the AI agent operates exclusively with targeted replacement blocks (Search-and-Replace / Unified Diffs), preserving the rest of the file unchanged and saving tokens.