Installing Graphify involves two steps: install the official graphifyy package, then register the platform-specific skill with Claude Code, Cursor, Aider, Codex, or another supported coding assistant. Once configured, Graphify gives the assistant a structured knowledge graph of your repository, helping it trace dependencies and reduce repeated file searches.
Does your AI coding assistant keep reopening the same files because it cannot see how your codebase fits together?
The 2025 Stack Overflow Developer Survey found that 84% of respondents either use or plan to use AI tools in development, while 51% of professional developers use them daily. Yet only 33% trust AI-generated output, compared with 46% who actively distrust it (Source).
This gap shows why coding agents need reliable structural context instead of depending only on repeated file searches and isolated code snippets.
Graphify addresses this problem by turning a repository’s code, documentation, schemas, and related files into a queryable knowledge graph.
Its official benchmark reported up to 71.5 times fewer tokens per query on a mixed 52-file corpus than reading the original files directly, although results vary by repository size and workflow (Source).
This Graphify installation guide explains the setup requirements, installation steps, graph generation process, update options, troubleshooting methods, and uninstallation procedure.
What Is Graphify?
Graphify is an open-source knowledge graph skill that helps AI coding assistants understand the structure of a codebase. It scans project files and maps relationships between:
Functions and classes
Modules and imports
APIs and schemas
Documentation and source code
Dependencies and architectural components
The generated graph gives coding assistants a structured view of the repository before they begin opening individual files.
This can help an assistant:
Trace dependencies across multiple files
Locate function, class, or module definitions
Understand major architectural areas
Assess which components may be affected by a code change
Graphify works alongside coding assistants rather than replacing them. It supports tools such as Claude Code, OpenAI Codex, Cursor, Gemini CLI, GitHub Copilot, and Aider.
Graphify Setup Requirements
Before installing Graphify, confirm that your system has a compatible Python version, a supported installation method, and access to an AI coding assistant.
Graphify requires:
Python 3.10 or later: Confirm your installed version using python --version.
A supported package manager: Graphify recommends uv, while pipx is the main alternative. Standard pip installation is also supported, although it may require manual PATH configuration.
Terminal access: You need PowerShell, Command Prompt, macOS Terminal, or a Linux shell to run the installation commands.
Access to the target repository: Graphify must be able to read the project files it will map.
A supported AI coding assistant: The official package supports Claude Code, OpenAI Codex, Cursor, Gemini CLI, GitHub Copilot, Aider, Kiro, OpenCode, Devin CLI, and more than 15 additional integrations.
Check the main requirements with:
python --version
uv --version
When using pipx, run:
pipx --version
Graphify also offers optional installation extras for PDFs, Office documents, videos, databases, watch mode, MCP, and external AI backends. API credentials are needed only when you configure a provider-specific backend such as Anthropic, OpenAI, DeepSeek, or Kimi.
How to Install Graphify Step by Step
The official PyPI package is graphifyy, while the CLI command you run after installation is graphify; use the exact package name when installing to avoid unaffiliated similarly named packages.
Step 1: Install uv or pipx
First, check whether either package manager is already available:
uv --version or pipx --version
When neither command works, install one using its official operating-system instructions. uv is the preferred option because it installs command-line tools in an isolated environment.
Step 2: Install the Official graphifyy Package
Install Graphify using the PyPI package graphifyy. You can use:
$ pip install graphifyy
or the project’s recommended installer, then verify the CLI with graphify --help.
After installation, run graphify install to register Graphify with your coding assistant.
Avoid similarly named packages. The correct package name is graphifyy, but all CLI commands begin with graphify.
Step 3: Verify the Graphify Installation
Run:
graphify --help
A successful installation displays the available Graphify commands.
When the terminal cannot find the command after a uv installation, run:
uv tool update-shell
Then reopen the terminal and run graphify --help again.
Configure Your Coding Assistant to Always Use the Graph
After building the graph, run the command for your coding assistant from the project directory. This installs the platform-specific integration so the assistant can consult Graphify during future coding tasks.
Use GRAPH_REPORT.md for a quick project overview and graph.html when you want to explore the graph visually.
Query the Graph for Architecture and Dependency Insights
Graphify supports commands such as:
graphify query "How does authentication work?"
graphify path "LoginController" "SessionStore"
graphify explain "PaymentService"
These commands help you:
Ask questions about the repository structure
Trace relationships between components
Explain the purpose of a specific entity
Assess dependencies before making code changes
How to Update a Graphify Knowledge Graph
Update the graph whenever files, dependencies, or repository structure changes.
Update Graphify Manually or With Git Hooks and Watch Mode
When your repository changes, refresh the graph with graphify . --update.
Use a full rebuild after major refactoring, branch restructuring, or changes to the folders included in the graph.
Graphify also supports automated updates:
Git hooks: Run the graphify hook install to update the graph after commits and branch changes. Use graphify hook status to check the hook, or graphify hook uninstall to remove it.
Watch mode: Run graphify . --watch in a separate terminal to detect supported code changes during development.
Watch mode is designed for fast structural updates. Changes to documents, images, or other semantically processed files may still require a manual update.
Common Graphify Setup Errors and How to Fix Them
Most setup issues relate to PATH configuration, Python environments, assistant registration, or outdated graph data.
Fix graphify: command not found and Python Environment Errors
When the terminal cannot find Graphify, update the shell PATH:
uv tool update-shell
Restart the terminal, then verify the installation:
graphify --help
When Graphify was installed with standard pip, the executable may be outside the active PATH or Python environment. Reinstalling it with uv or pipx can reduce these conflicts:
uv tool install graphifyy
Remember that graphify . runs in the terminal, while /graphify . is used inside a supported AI coding assistant.
Fix Missing Integrations, Output Files, or Outdated Graphs
When the /graphify skill is unavailable, register it again and restart the coding assistant:
graphify install
When Graphify does not generate output files:
Confirm that you are in the correct repository directory.
Check that Graphify can read the project files.
Review the terminal for permission or dependency errors.
Confirm that the command completed successfully.
When the existing graph no longer matches the repository, refresh it with:
graphify . --update
How to Uninstall Graphify
To uninstall Graphify, run graphify uninstall. Use --purge only if you also want to remove the generated graph files.
This keeps the generated graphify-out/ directory. To delete those files as well, run:
graphify uninstall --purge
You can also remove a specific integration without affecting the others. For example:
graphify claude uninstall
Next, uninstall the graphifyy package using the package manager originally used for installation:
uv tool uninstall graphifyy or pipx uninstall graphifyy
For a standard pip installation, run:
pip uninstall graphifyy
Use --purge only when you no longer need the generated graph, report, or visualization files.
Graphify can be installed in a few steps using the correct graphifyy package, a compatible Python environment, and the appropriate coding-assistant integration.
After setup, generate the first knowledge graph, verify that the output files were created, and choose an update method that matches your development workflow. Most setup issues can be resolved by checking the system PATH, active Python environment, repository permissions, or assistant registration.
Following these steps gives your coding assistant a structured view of the repository and helps it navigate project relationships more efficiently.
Give Your AI coding Agent a Real Memory Layer
Install Harmony and reduce the time, tokens, and context loops wasted on finding the right files.
Graphify turns code, documentation, schemas, and related project files into a queryable knowledge graph. AI coding assistants can use it to understand repository structure and trace dependencies more efficiently.
Is Graphify Free and Open Source?
Yes. Graphify is open source under the MIT License and can generally be used for personal or commercial projects, subject to the license terms.
What Are the Best Alternatives to Graphify?
Alternatives depend on the use case. Harmony MCP is better suited to teams that need ranked context retrieval, persistent memory, and token-controlled context delivery beyond repository mapping.
Does Graphify Actually Save Tokens?
It can reduce token usage by letting assistants query a compact graph instead of repeatedly reading full files. Graphify reports up to 71.5× fewer tokens in its own benchmark, though real results vary by repository and workflow.
Use Graphify to give coding agents reliable structural codebase context.
Verify Python 3.10+, terminal access, and repository read permissions first.
Install with uv or pipx, then pip install graphifyy carefully.
Run graphify install and graphify --help to confirm integration success.
Written by
Paul Dhaliwal
Founder & Chief Executive Officer
Paul Dhaliwal is a tech innovator and Founder of CodeConductor, an open-source no/low-code platform. With 10+ years of experience in AI and scalable development, Paul focuses on crafting intelligent solutions that drive real-world value. A firm believer in the mantra "Eat, Sleep, Code, Repeat," he balances his passion for software with a love for travel and family.
⚡
Build your app
No coding. No designers. Just describe what you want and watch AI build it.
