Graphify: Step-by-Step Setup Guide
1. Prerequisites
Make sure you have:
● Python 3.10+ installed
● pip available
● An AI coding assistant (one of):
○ Claude Code
○ Codex
○ Cursor
○ Gemini CLI
○ GitHub Copilot CLI
○ VS Code Copilot Chat
2. Install Graphify
Run:
pip install graphifyy && graphify install
⚠️ Important:
● The correct package name is graphifyy (double y)
● Other similarly named packages are not official
3. Platform-Specific Setup
Depending on your tool:
Claude Code
graphify install
Cursor
graphify cursor install
Codex
graphify install --platform codex
Gemini CLI
graphify install --platform gemini
GitHub Copilot CLI
graphify install --platform copilot
VS Code Copilot Chat
graphify vscode install
4. Run Graphify on Your Codebase
Navigate to your project folder and run:
/graphify .
For Codex:
$graphify .
This will generate:
graphify-out/
├── [Link]
├── GRAPH_REPORT.md
├── [Link]
└── cache/
5. Make Graph Always Available to AI (Recommended)
Run:
graphify claude install
Or equivalent for your platform:
graphify cursor install
graphify codex install
graphify gemini install
This ensures:
● AI reads graph before searching files
● Better architectural understanding
● Reduced token usage
6. Ignore Unnecessary Files
Create a .graphifyignore file:
node_modules/
dist/
vendor/
*.[Link]
7. Query Your Graph
Examples:
graphify query "show auth flow"
graphify query "what connects X to Y"
graphify path "A" "B"
graphify explain "SomeConcept"
8. Give context to Agent
Prompt (Paste as it is in agent, just update the GRAPHIFY_OUT_PATH) :
You have access to a Graphify knowledge graph at:
<GRAPHIFY_OUT_PATH>
For every question:
1. First run:
graphify query "<user question>" --graph <GRAPHIFY_OUT_PATH>/[Link]
2. Use ONLY the returned graph context to answer.
3. If the graph context is insufficient, then (and only then) read specific files.
Rules:
- Do NOT scan the entire codebase
- Do NOT load full files unless absolutely necessary
- Prefer relationships, dependencies, and paths from the graph
- Cite source files mentioned in the graph output when possible
How to Keep Growing the Graph
Graphify becomes more powerful as your corpus grows. To continuously improve it:
● Re-run periodically
/graphify .
This updates the graph with new files and changes.
● Add more context sources Include:
○ Docs (.md, .txt)
○ PDFs (papers, specs)
○ Screenshots / diagrams
○ Notes / raw ideas
● Use add for external knowledge
graphify add <url>
Example: papers, blogs, tweets, videos
● Keep raw knowledge folder Maintain something like:
/raw
Drop everything there → Graphify connects it
● Use update instead of full rebuild (faster)
graphify update ./
● Think in signals, not files The graph improves when you add:
○ reasoning
○ comments
○ architectural decisions
Over time, your graph becomes a persistent memory layer for all agents, not just a one-time
analysis.
Reference
Based on official repo instructions. Source: [Link]