Large AI coding assistants are impressive—but once your project grows, they start spending a lot of time searching files, consuming tokens, and sometimes missing important context.

I recently started using Graphify with OpenCode, and it made a noticeable difference. Instead of scanning raw files every time, OpenCode can query a knowledge graph of my project, making responses faster and much more accurate.

In this guide, I'll show you exactly how I set up Graphify on Windows with a Next.js project.

What is Graphify?

Graphify is an open-source tool that analyzes your codebase and builds a local knowledge graph.

Instead of treating your project as thousands of files, it understands:

• 📦 Classes

• ⚙️ Functions

• 📂 Modules

• 🔗 Imports

• 🕸 Relationships between components

Everything runs locally, so your source code never leaves your machine.

It also integrates with AI coding assistants like:

• OpenCode

• Cursor

• Claude Code

• Other MCP-compatible AI tools

Prerequisites

Before getting started, make sure you have:

• Windows

• Python 3.10+

• A Next.js (or any supported) project

• OpenCode CLI installed

• PowerShell

Step 1: Install Python

Download Python from:

👉 https://www.python.org/downloads/

During installation, don't forget to enable:

✅ Add python.exe to PATH

After installation, verify:

python --version

Example:

Python 3.14.x

Step 2: Install Graphify

Install Graphify using pip:

pip install graphifyy

Note: The package name is graphifyy (double y), while the command is simply graphify.

If the command isn't recognized, install it using pipx instead (recommended):

python -m pip install pipx

python -m pipx ensurepath

Restart PowerShell, then run:

pipx install graphifyy

Verify the installation:

graphify --version

Step 3: Connect Graphify to OpenCode

Navigate to your project:

cd your-project

Run:

graphify opencode install

This automatically:

• Updates your AGENTS.md

• Installs the Graphify plugin

• Instructs OpenCode to prefer the knowledge graph before searching files

Step 4: Build Your First Knowledge Graph

Start OpenCode:

opencode

Inside OpenCode, run:

/graphify .

Graphify scans your project and generates:

graphify-out/
├── graph.html
├── GRAPH_REPORT.md
└── graph.json

Files Explained

graph.html

Interactive visualization of your project architecture.

GRAPH_REPORT.md

A readable architecture report generated from your codebase.

graph.json

The underlying knowledge graph used by AI assistants.

The initial scan may take a few minutes depending on your project size.

Step 5: Make It Always-On (Recommended)

Run the install command once more:

graphify opencode install

This reinforces OpenCode's behavior so it consults the knowledge graph before scanning raw source files.

Add to .gitignore

graphify-out/manifest.json
graphify-out/cost.json
graphify-out/cache/
graphify-out/20*/

How to Know It's Working

Ask OpenCode something architecture-related:

How does authentication work in this project?

If you notice it reading:

graphify-out/GRAPH_REPORT.md

or referring to concepts like:

• Communities

• Dependencies

• Graph nodes

• God nodes

then Graphify is working correctly.

You can also be explicit:

Use the Graphify knowledge graph to explain the application's data flow.

Useful Commands

Query your project:

graphify query "How does authentication connect to the database?"

Explain a component:

graphify explain "AuthService"

Update the graph after code changes:

graphify . --update

Why I Like Graphify

The biggest improvement isn't just visualization—it's how much better AI assistants understand large projects.

Instead of repeatedly searching hundreds of files, OpenCode can reason over a structured graph of your application.

I've noticed:

• ⚡ Faster responses

• 🎯 Better architectural understanding

• 💰 Lower token usage

• 🧠 More accurate code explanations

For medium-to-large codebases, it's a simple upgrade that significantly improves the AI coding experience.

Final Thoughts

Setting up Graphify took me less than 20 minutes, and most of that time was installing Python.

If you're using OpenCode, Cursor, or Claude Code on a growing codebase, I highly recommend giving Graphify a try.

It helps your AI understand your project the way you understand it—as a connected system instead of a collection of files.

Happy coding! 🚀

Useful Links

Graphify GitHub

https://github.com/graphify-ai/graphify

Python Downloads

https://www.python.org/downloads/

OpenCode

https://opencode.ai