Skip to main content
AGIT

Code Context Management

Capture the reasoning context (Why/How) alongside your code changes (What). AGIT creates a "Neural Graph" parallel to Git's commit graph, giving you a complete history of not just what changed, but why it changed and how the decision was made.

Get StartedView on GitHub
$ agit init
Initialized AGIT in /your-project/.agit
Created: .cursorrules, CLAUDE.md, .windsurfrules
 
$ agit record "Planning to refactor auth module"
Recorded thought to staging area
 
$ agit commit -m "Refactor auth module"
Created git commit: e8d4f1a
Created neural commit: a3f7b2c
Summary: Intent: Refactor auth module. Plan: Add try/catch around token validation.

Why AGIT?

Seamless Integration

Works with Cursor, Claude Code, Windsurf, and other AI coding assistants. Your AI automatically logs thoughts via MCP.

Zero Configuration

Just run `agit init` and you're ready. Auto-generates config files for all major AI editors.

Neural Graph

Creates a parallel graph of reasoning alongside Git's commit graph. Query the "why" behind any change.

Git Compatible

Built on libgit2, works alongside your existing Git workflow. Not a replacement—an enhancement.

How It Works

The "Seamless Echo" Strategy

Unlike traditional tools that call LLMs, AGIT inverts the flow. Your AI editor pushes context to AGIT via MCP:

  1. User asks: "Fix the auth bug"
  2. AI logs intent: agit_log_step(role="user", ...)
  3. AI plans: "I'll add a try/catch block"
  4. AI logs plan: agit_log_step(role="ai", ...)
  5. User commits: agit commit -m "Fix auth bug"
  6. AGIT synthesizes: Links Intent + Plan to git commit

Quick Install

# macOS
$ brew tap agit-stuff/agit && brew install agit
 
# Linux / CI
$ curl -fsSL https://agit.dev/install.sh | bash
 
# Windows (PowerShell)
> scoop bucket add agit https://github.com/agit-stuff/agit; scoop install agit
All Installation Options