Graphify: a free knowledge graph for any codebase
Free, open-source Claude Code skill that turns any codebase into a knowledge graph. Ask questions in plain English, runs 71x cheaper, works with a local mode.
What it is
Imagine handing a very thorough intern all your documents, screenshots, and files — every folder, every PDF, every note — and asking them to draw you a map of how everything connects. Then imagine being able to ask that map questions. That's roughly what Graphify does.
You type /graphify inside Claude Code (the AI coding assistant from Anthropic), and it reads through your project. It comes back with an interactive diagram, a report in plain English, and a file you can search. It spots the parts of your project that do too much, finds hidden connections you didn't notice, and explains why things were built the way they were.
Why it's interesting
The clever part is efficiency. Reading through a big project file by file is expensive — it uses a lot of the AI's attention. Graphify maps it first, so when you ask a question, the AI only needs to look at the relevant corner of the map instead of rereading everything. The creators claim it uses 71 times fewer resources per question.
The newest version also does some of its thinking entirely on your own computer, which means no extra costs and nothing leaving your machine.
Worth thinking about
If you work with an agency, a developer, or any team that builds things — ask them how they currently document decisions. A tool like this could answer the question you've always wanted to ask: why is it built this way?
Quick glossary
Knowledge graph — a diagram that shows not just a list of things, but how they relate to each other. Like a mind map, but generated automatically.
Token — the unit of "attention" an AI uses to read text. Fewer tokens means faster, cheaper responses.
Open-source — software whose blueprint is public. Anyone can read it, use it, or improve it.
Local / on-device — when the AI works on your own computer instead of sending data to a server somewhere else.
Try it
Map your project in under two minutes — Claude Code does the work.
- Install Claude Code:
npm install -g @anthropic-ai/claude-code - Add the Graphify skill:
claude skill add safishamsi/graphify - Open your project folder in the terminal and run
claude - Type
/graphifyand let it map the codebase - Ask plain-English questions like why is this built this way?
Frequently asked
Does Graphify work with a Claude free account?
Yes. Graphify runs as a Claude Code skill, so any plan that lets you use Claude Code works. On the free tier you'll hit message limits faster, which is exactly why Graphify's 71x efficiency gain matters — you burn through fewer tokens per question. For heavier projects, switch on the local mode and the mapping happens on your machine without touching the Claude quota.
Do I need a paid Anthropic plan to use Graphify?
No paid plan is required to try it. Graphify is free and open-source. What you need is Claude Code installed, and any Claude account it accepts — free, Pro, or API. A paid plan removes rate limits, but the skill itself costs nothing.
Does Graphify send my code to Anthropic?
Only the parts Claude Code already sees when you chat with it. Graphify doesn't add a second pipeline — it just gives Claude a smarter map so it asks for less. The newest version also supports a local mode where most of the mapping runs entirely on your computer.
How is Graphify different from just asking Claude Code about my project?
Claude Code reads files on demand, which gets expensive on big projects and often misses cross-file connections. Graphify builds the whole knowledge graph once, then lets Claude query the map. You get faster answers, lower cost, and it notices relationships between files that a file-by-file read would miss.