2aronS

2aronS / agent-git

Public

Git-based knowledge management system for AI agents

ai-agents git knowledge-base knowledge-management rust
11
0
89% credibility
Found May 21, 2026 at 11 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

agent-git is a versioned knowledge management system for AI coding assistants. Think of it as giving your AI agent a searchable, version-controlled memory that travels with your project. Just like git tracks changes to code, agent-git tracks facts, decisions, and context that your AI needs to remember. You can connect it to AI coding tools so your assistant automatically learns about your project, or use it to ask questions about past decisions. The same fact stored by different users or sessions gets the same ID automatically, so knowledge deduplicates itself. It works entirely locally with no external services required.

How It Works

1
💬 You struggle with AI agents forgetting everything

You use Claude Code or Cursor for coding, but each session starts fresh - your agent has no memory of previous conversations or decisions.

2
🤔 You wish your AI could remember things

You've been manually copying notes and context into each chat, hoping your AI assistant remembers important details from last week.

3
✨ You discover a smart notebook that never forgets

agent-git works like a shared notebook your AI can write to and read from, creating a searchable graph of everything important.

4
📝 You create a project memory

You run one simple command and a hidden .agent-git folder appears in your project - your AI's new knowledge home.

5
🎓 Your AI learns your project

You can feed documents, write notes, or let your AI automatically save important decisions it discovers while working.

6
Choose how you want to use it
🔌
Connect directly to your AI tool

One command plugs agent-git into Claude Code or Cursor - your AI starts using your knowledge automatically.

🔍
Ask questions anytime

You can search your project's memory whenever you want: 'What did we decide about the API design?'

🌐
Share across your whole team

Push and pull your project knowledge like code - everyone on your team has the same AI memory.

🎉 Your AI becomes genuinely smart

Your assistant now remembers your coding conventions, design decisions, and project context across every session - finally!

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 11 to 11 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is agent-git?

Agent-git is version control for AI agent memory. Built in Rust, it gives your coding assistants (Claude Code, Cursor, Gemini CLI, and any MCP-compatible tool) a persistent, branchable knowledge base that travels with your codebase. Instead of starting every session from zero, agents can store facts, decisions, and context, then retrieve exactly what matters for the task at hand. The retrieval engine combines semantic vector search, keyword matching, and graph traversal in a single query, and reports exactly what it found, what it skipped, and how many tokens were consumed. It ships as a single binary with no external database required.

Why is it gaining traction?

The killer feature is branching and merging for knowledge. Just like git lets you experiment with code on a branch and merge when ready, agent-git lets agents explore approaches in isolation and merge conclusions back. This is something no other agent memory system offers. The transparency is also notable: every query returns candidates seen, tokens used, and dropped results, so you know exactly what your agent did and did not consider. The benchmarks show strong retrieval numbers against mem0 and MemPalace across six public datasets, with reproducible methodology and raw artifacts published.

Who should use this?

Devs who live in Claude Code or Cursor and are tired of re-explaining context each session. Teams maintaining multiple AI-assisted projects where you want consistent agent behavior across the codebase. Anyone building agents that need to remember decisions across sessions without relying on vendor-hosted memory services. The sync commands (push, pull, clone) make it viable for CI pipelines where automated agents can commit findings and developers pull them locally.

Verdict

Agent-git is a well-engineered solution to a real problem, with solid Rust code, thoughtful documentation, and benchmarks that hold up to scrutiny. The credibility score of 0.9% and 11 stars reflect an extremely young project (v0.1.0, pre-1.0) rather than any technical weakness. If you use AI coding assistants professionally, this is worth a weekend experiment. The one-binary deployment and local-first design mean minimal risk to try. Just monitor the changelog before updating in production.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.