aig — Version Control for the AI Age

Git tracks what changed. aig tracks why.

aig is an intent-based version control layer that sits on top of git. It captures the goal behind every change, the AI conversation that shaped it, and the structural impact on your codebase — so your history tells a story, not just a diff.

Every aig repo is a valid git repo. Your existing tools, CI, and workflows keep working unchanged.

See the difference

git log gives you this:

43ed456 Initial commit
fd990ed Add auth module
7564dca Fix tests
6bdca12 Refactor middleware

aig log gives you this:

[a3f1b2c4] Add JWT authentication (done)
         3 checkpoint(s) | 2 h 15 min
           (b7c2d3e4) Token generation and validation
                     + added generate_token (auth.py)
                     + added validate_token (auth.py)
           (d4e8f9a0) Auth middleware integration
                     ~ modified setup_routes (routes.py)
           (f1a2b3c4) Test coverage
                     + added test_auth_flow (test_auth.py)

Four commits become one intent. You see the narrative, not the noise.

git blame gives you a name and a date. aig why gives you the full story:

$ aig why src/auth.py:42

  Intent:     [a3f1b2c4] Add JWT authentication
  Checkpoint: Token generation and validation
  Commit:     b7c2d3e4
  Time:       2026-04-14 01:06

  Semantic changes:
    + added `generate_token`
    + added `validate_token`

  Conversation notes:
    - Chose HS256 over RS256 for simplicity in single-service deployment

Quick start

# Install (Linux / macOS)
curl -fsSL https://raw.githubusercontent.com/saschb2b/ai-git/main/scripts/install.sh | sh

# Install (Windows)
irm https://raw.githubusercontent.com/saschb2b/ai-git/main/scripts/install.ps1 | iex

# Try it on any existing git repo
cd your-project
aig init --import        # initialize + import git history in one step
aig log                  # browse intents instead of flat commits
aig why src/app.py:42    # trace any line to its intent

How it works

# 1. Declare what you're working on
aig session start "Add user authentication"

# 2. Work normally — edit files, use AI, iterate
# 3. Checkpoint when ready (message auto-generated from code changes)
aig checkpoint

# 4. Add reasoning (optional but valuable)
aig conversation add "Chose bcrypt over argon2 for compatibility"

# 5. End the session
aig session end

# 6. Share with your team
git push && aig push

Features

Feature	Command	What it does
Intent tracking	aig session start/end	Declare goals before writing code
Smart checkpoints	aig checkpoint	Git commit + semantic analysis, auto-generated messages
Semantic diff	aig diff --semantic	AST-level changes across 11 languages
Line provenance	aig why file:line	Trace any line to its intent and conversation
LLM explanations	aig why --explain	Natural-language synthesis of why code exists
Trust scoring	aig trust [file]	Track human vs AI-authored code regions
Interactive review	aig review --tui	Terminal UI for navigating intents
Git history import	aig import	Retrofit intents onto existing repos (incremental)
AI conversation capture	aig capture	Auto-capture from Claude Code or any AI tool
Remote sync	aig push / pull	Share intent metadata via git notes
Git hooks	aig hooks install	Auto-checkpoint, auto-session, auto-sync
File watching	aig watch	Auto-checkpoint after quiet periods
Releases	aig release <tag>	Tag a release grouping intents since the last one
Changelog	aig changelog	Auto-generate release notes from intents
Portable backup	aig export	Bundle .aig metadata for backup or migration

Installation

One-line install (recommended)

Linux / macOS:

curl -fsSL https://raw.githubusercontent.com/saschb2b/ai-git/main/scripts/install.sh | sh

Windows (PowerShell):

irm https://raw.githubusercontent.com/saschb2b/ai-git/main/scripts/install.ps1 | iex

Auto-detects your platform, downloads the latest release, and adds aig to your PATH.

From source

cargo install --git https://github.com/saschb2b/ai-git.git aig-core

Manual download

Browse all releases for pre-built binaries (Linux x86_64, macOS aarch64, Windows x86_64).

Build locally

git clone https://github.com/saschb2b/ai-git.git
cd ai-git
cargo build --release
# Binary at target/release/aig (.exe on Windows)

Requirements: Rust 1.75+ and an existing git repository. Node.js 20+ is optional (needed for LLM-powered import and TUI review).

Semantic diff: 11 languages

TypeScript/JS, Python, Rust, Go, Java, C#, C++, Ruby, PHP, Kotlin, Swift. All other languages fall back to line-based diffing automatically.

Documentation

• Getting Started — install and try it in 60 seconds
• Daily Workflow — how aig fits into your routine
• CLI Reference — all 21 commands with flags and examples
• Migration Guide — import existing git repos
• Roadmap — what's shipped, what's next
• RESEARCH.md — the full vision (~5,000 words)
• TECH_STACK.md — technology choices and rationale

Repository structure

crates/aig-core/          Rust CLI + library (21 commands)
crates/aig-treesitter/    Semantic diff engine (11 languages, tree-sitter)
packages/aig-llm/         LLM integration (TypeScript, Anthropic Claude)
packages/aig-tui/         Interactive TUI (TypeScript, React/Ink)
docs/                     VitePress documentation site

License

MIT