Memento: Git Extension for AI Session Tracking

Memento is a Git extension that records AI coding sessions with commits, enhancing traceability in AI-assisted development. It stores markdown conversations as Git notes, allowing developers to audit AI contributions. The target audience includes developers using AI coding assistants like Codex and Claude, providing a benefit of improved code provenance and reproducibility.

How It Works

Memento captures AI sessions upon commit, storing them as Git notes (full transcripts in refs/notes/memento-full-audit, summaries in refs/notes/commits). It supports extensible AI provider integration and configurable summary skills, documenting AI's role without disrupting Git workflows.

Quick Start & Requirements

Install via curl -fsSL https://raw.githubusercontent.com/mandel-macaque/memento/main/install.sh | sh. Initialize with git memento init and provider setup. Core commands include git memento commit <session-id> -m "message", git memento amend <session-id> -m "message", git memento push, and git memento notes-sync. Building from source requires .NET SDK 10+ and native toolchain for NativeAOT. AI provider CLIs must be configured.

Highlighted Details

• GitHub Actions: Reusable marketplace actions for commit comments, CI gates (audit), and automatic note carry-over to merge commits (merge-carry).
• NativeAOT Compilation: CLI built with .NET NativeAOT for single, self-contained executables (macOS, Linux, Windows).
• Extensible Provider Support: Designed for multiple AI assistants beyond Codex/Claude.
• Note Management: Supports full logs and customizable summaries.
• Rewritten Commit Preservation: Commands (notes-rewrite-setup, notes-carry) maintain notes across rebases/amendments.

Maintenance & Community

Provides a GitHub Actions integration for CI/CD pipelines. Release automation ensures timely updates. No specific contributors or sponsorships detailed.

Licensing & Compatibility

The project's license is not explicitly stated, creating ambiguity for commercial use or closed-source integration. NativeAOT build targets cross-platform compatibility (macOS, Linux, Windows).

Limitations & Caveats

Functionality depends on correctly installed/configured external AI provider CLIs. Building from source requires specific .NET SDK versions and native toolchain setup. Absence of a declared license is a critical limitation.