💡 What is this? Graphify is a free, open-source tool that builds a knowledge graph of your entire codebase — so Claude Code navigates relationships instead of re-reading files every session. The result: up to 70% fewer tokens consumed.
🧠 The Problem It Solves
| Problem | Impact |
| Claude re-reads every file on each new session | Thousands of tokens burned before you ask a question |
| Large codebases = instant token drain | Most users exhaust their plan in 2 days |
| $20 plan feels underpowered | Limits creativity and productivity |
⚙️ How Graphify Works
Step 1 — Run it once on your codebase
Graphify scans every file and function across your entire project.
Step 2 — It builds a full knowledge graph
Every connection and relationship between files, functions, and modules is mapped.
Step 3 — Claude navigates, not reads
In future sessions, Claude traverses the graph instead of re-reading raw files — slashing token usage by up to 70x.
🚀 Setup Guide
📦 Installation & First Run
- Open the GitHub link: https://github.com/safishamsi/graphify
- Clone the repo into your project root:
git clone https://github.com/safishamsi/graphify
cd graphify- Run the indexer on your codebase:
python graphify.py --path /your/project/path- Point Claude Code to the graph — follow the repo README for the Claude integration step
- Start a new Claude Code session — token usage should drop dramatically from the first query
✅ Verifying It's Working
- Compare token counts before and after on a sample session
- Claude should answer structural questions ("where is X called?") without reading source files
- First session after indexing will still consume setup tokens — savings appear from session 2 onward
💬 Prompts to Use with Graphify
🗺️ Explore Your Codebase via the Graph
Using the knowledge graph, give me an overview of how [feature or module name] is structured. What are the key files, functions, and dependencies involved? Don't read files — navigate the graph.🔍 Find Relationships Between Components
Without reading source files, use the graph to tell me: what calls [function name], what does it depend on, and where does its output go?🛠️ Targeted Debugging with Minimal Tokens
I'm seeing an issue with [describe bug]. Using the graph, identify the most likely files and functions involved. Only read a file if you can't answer from the graph alone.📊 Token Savings at a Glance
| Plan | Before Graphify | After Graphify |
| Claude Pro ($20/mo) | Exhausted in ~2 days | Lasts the full month |
| Claude Max ($100/mo) | Needed for large projects | Replaced by $20 plan |
| Token burn per session | High (full re-read) | ~70% lower |