changAIs

Finally know exactly what your AI just did to your code.

changais.com · Docs · Install · Roadmap · VS Code Marketplace · Open VSX

changAIs is a VS Code extension for reviewing changes made by AI coding tools. It shows changed files in a diff-focused review flow and attaches AI-generated notes to the lines they describe, so you can inspect what changed before committing.

changAIs helps you see and organize what an AI tool produced, but it cannot make that output correct. Treat every AI note, explanation, and suggested change as something to verify before you accept or commit it.

What changAIs Does

• Shows files changed by an AI coding session.
• Opens changed files in a diff view.
• Displays AI notes near the relevant changed lines.
• Lets you reply to notes while reviewing.
• Lets you mark notes as checked or resolved as you work through them.
• Helps you understand report files generated by changAIs-compatible workflows.

Supported Editors

changAIs supports Visual Studio Code and compatible VS Code-based editors that can install extensions from a marketplace or from a .vsix file.

Use the Visual Studio Marketplace for VS Code, the Open VSX Registry for VSCodium and compatible editors, or the .vsix install flow documented in Install.

Getting Started

Follow the full guide in Getting Started.

1. Install the Extension

Install changAIs from the extension marketplace for your editor:

• VS Code: Visual Studio Marketplace.
• VSCodium and compatible editors: Open VSX Registry.

If you are installing from a downloaded package, open the Command Palette and run Extensions: Install from VSIX....

2. Open a Project

Open the folder that contains the code you want to review. changAIs works best when the project already has changed files from an AI coding session.

3. Confirm changAIs Is Available

Open the Extensions view and confirm that changAIs is installed and enabled for the current workspace. You can also open the Command Palette and type changAIs to see available actions.

4. Set Up Your AI Tool

Before an AI assistant can generate changAIs reports automatically, add changAIs report instructions to the project. The extension includes setup commands for Claude Code, GitHub Copilot, Cursor, Windsurf, Aider, Gemini CLI, OpenCode, and Qwen Code.

Run the setup command that matches your AI tool, then ask the AI to make changes as usual. The generated instructions tell the AI to create or replace .changais/report.json, add notes for touched paths, and keep line numbers accurate so changAIs can pin notes to the diff.

5. Load or Generate a Report

Reports connect your changed files with AI-generated notes. A report may come from:

• A changAIs-compatible AI workflow.
• A manually selected report file.
• A generated report from a supported changAIs workflow.

To trigger report generation manually, run changAIs: Generate Report. CLI-based tools run in a terminal; tools that need manual prompt handoff show a copyable prompt.

6. Review Before You Commit

Work through each changed file, compare the diff, and inspect notes pinned to the relevant lines. Use checked or resolved state to track what you have already reviewed.

When you finish, commit only the changes you understand and accept.

Reports

Read the full guide in Reports.

Reports must match the project and file state you are reviewing. If files were moved, deleted, or changed after the report was created, notes may not line up with the current diff.

The normal report workflow is:

AI makes changes
      |
      v
AI writes .changais/report.json
      |
      v
Open the changAIs review surface
      |
      v
Review changed files, notes, replies, and checked state
      |
      v
Prepare AI follow-up prompt if needed
      |
      v
Review again, or commit when satisfied

Sending Questions Back to AI

Replies create a feedback loop for the current report:

1. Review a note in changAIs.
2. Add a reply such as question, agree, disagree, or will-fix.
3. Run changAIs: Prepare AI Follow-up Prompt, or click Prepare AI Prompt in the changAIs review toolbar.
4. Review .changais/prompt.md, then give it to your AI tool or pass it through your own workflow.
5. The AI should read .changais/prompt.md, .changais/report.json, and .changais/replies.json, act on the feedback, and update .changais/report.json with fresh notes.

changAIs does not send your code or replies to an AI service automatically. It writes a local .changais/prompt.md file. You decide whether to paste that prompt into your AI assistant or pass it through your own workflow.

Report Format

Read the full guide in Report Format.

AI tools generate a JSON report at .changais/report.json. The format is intentionally simple so you can create changAIs reports from scripts, CI jobs, or custom AI coding workflows.

{
  "version": "1.0",
  "generatedAt": "2026-06-08T12:00:00Z",
  "notes": [
    {
      "id": "note-auth-guard",
      "file": "src/auth/session.ts",
      "line": 42,
      "endLine": 48,
      "type": "warning",
      "title": "Check unauthenticated behavior",
      "content": "The new guard changes how expired sessions are handled. Confirm redirects and API responses still match the product flow.",
      "checked": false
    }
  ]
}

Required fields are version, notes, and each note's id, file, line, and content. Use 1-based line numbers from the current file content and paths relative to the workspace root opened in VS Code.

Review Notes

Read the full guide in Review Notes.

Start with the changed-files list, then open each file and compare the before and after state in the diff view.

AI notes may explain intent, flag risk, suggest a follow-up, or document a decision made during generation. Treat notes as prompts for review, not guarantees.

Use replies to capture your own review decisions or follow-up questions.

Use checked or resolved state to track progress through the review:

• Checked means you looked at the note.
• Resolved means the note no longer needs action.

Features

• Diff-focused review for AI-generated code changes.
• Notes pinned to changed files and changed lines.
• Note types: suggestion, warning, and info.
• Replies: agree, disagree, question, and will-fix.
• Checked or resolved state for review progress.
• Report generation and AI follow-up prompt handoff.
• Multi-line notes with endLine support.
• Filter, search, and jump-to-note workflows.
• Auto-refresh when reports change.
• Auto-clear after commit.
• Markdown export.

Pro Features

• Configurable report path.
• Report history and report comparison.
• Merge multiple reports into one view.
• HTML, CSV, and PDF export.
• Analytics tracking.
• Commit with AI.

Commands

Configuration

Custom Reply Templates

The changais.replyTemplates setting lets you add project-specific quick-reply buttons to note hover cards. Each template requires a label and type, with optional text for the reply body:

"changais.replyTemplates": [
  { "label": "Out of scope", "type": "disagree", "text": "This is out of scope for this PR." },
  { "label": "Ticket created", "type": "will-fix", "text": "Created a follow-up ticket for this." }
]

Valid type values are agree, disagree, question, and will-fix.

FAQ

Read the full FAQ.

Does changAIs send my code anywhere?

No. changAIs runs inside your editor and reads local project and report data for the review workflow. It does not upload your code as part of normal use.

Is changAIs a code review replacement?

No. changAIs helps you inspect AI-generated changes, but you still decide what is correct, safe, and ready to commit.

What should I check before committing AI-generated changes?

Review the diff, read the AI notes, verify risky behavior with tests or manual checks, and commit only the changes you understand and accept.

Roadmap

Read the full Roadmap.

Planned areas include more AI tool integrations, team workflows, test and lint actions, review time tracking, AI usage and cost tracking, richer review checklists, and local review history.

License

changAIs Proprietary License. See the packaged LICENSE.md file.