codebreaker77

codebreaker77 / Fullerenes

Public

Persistent memory for AI coding agents.

14
1
100% credibility
Found Apr 28, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Fullerenes creates a local map of a codebase's structure for AI coding agents to query efficiently, reducing token usage and improving context awareness.

How It Works

1
🔍 Hear about Fullerenes

You discover a handy helper that gives AI coding buddies a smart memory of your entire project.

2
🚀 Start in your project

From your project's main folder, you launch it with one easy command and let it explore your files.

3
🗺️ Magic map appears

It builds a clear picture of all your functions, how they connect, and key starting points – everything feels organized.

4
📝 Get helper notes

Special guide files pop up for your AI, summarizing the big picture so it jumps right in.

5
❓ Ask away

You chat with it in plain words like 'show me user login' and get just the right snippets without overload.

6
👀 Stay fresh

Switch on watch mode to automatically refresh the map whenever you tweak your code.

🎉 AI gets super smart

Your AI helper now zooms to the exact code spots, works faster, and nails changes without guesswork.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 14 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 Fullerenes?

Fullerenes builds a local SQLite knowledge graph from your codebase, letting AI coding agents query files, symbols, callers, and impact zones without reloading raw context every session. Run `npx fullerenes init` in TypeScript, Python, Rust, Go, or Java projects to index incrementally, then query via CLI like `fullerenes query "how does auth work"` or connect over MCP for tools like `query_codebase` and `predict_impact`. It generates AGENTS.md and CLAUDE.md files with graph-grounded summaries, plus watch mode keeps everything fresh—perfect for github persistent memory in large repos.

Why is it gaining traction?

Unlike raw file dumps that burn tokens, Fullerenes delivers compact, budgeted outputs (94% token savings in benchmarks) with natural-language retrieval and no external LLM calls. MCP server plugs straight into Claude or Cursor for agent-native tools, while CLI stats and subgraph queries give devs instant repo insight. Local-first persistence beats hosted graph services, echoing fullerenes properties like robust structure in nanotubes and carbon nanostructures.

Who should use this?

AI-heavy devs on monorepos using Claude Desktop or Cursor for refactors, where agents struggle with entry points and blast radius. Teams building fullerenes c60-inspired persistent empires on github persistent storage, tired of context loss between sessions. Not for tiny scripts—best for mid-to-large codebases needing caller graphs and incremental syncs.

Verdict

Try it if you're deep in AI-assisted coding; solid docs and CLI make onboarding fast despite 14 stars and 1.0% credibility signaling early maturity. Polish tests and add JS parser tweaks for broader appeal—promising persistent memory github tool otherwise.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.