Skip to content
140.0k
Log inGet started

BlogGuides

Install Graphify: graphifyy with uv, or the hosted app

· 4 minute read

Graphify is a code knowledge graph that your AI coding agent queries instead of grepping, and there are two ways to install Graphify. You can install the open-source Graphify CLI from PyPI and build graphs on your machine. Or you can use hosted Graphify at app.graphify.com, which needs no install beyond connecting your agent.

This guide covers both. It takes about five minutes either way.

What do I need before I install Graphify?Link to What do I need before I install Graphify?

For the Graphify CLI you need Python 3.10 or newer and a package tool that installs command-line apps, ideally uv or pipx. You also need an AI coding assistant such as Claude Code, Cursor or Codex. For hosted Graphify you need a GitHub or GitLab account and an assistant that supports remote MCP servers.

If you do not have uv yet, the uv installation guide has a one-line installer for macOS, Linux and Windows.

Install the Graphify CLI with uvLink to Install the Graphify CLI with uv

The package on PyPI is graphifyy, with two y's. The command it installs is graphify.

bash
uv tool install graphifyy

uv puts the tool in its own environment, so it does not clash with your project's dependencies. If you prefer pipx, pipx install graphifyy does the same job.

If your shell then says graphify: command not found, the tool directory is not on your PATH yet. Fix it once:

bash
uv tool update-shell

Open a new terminal and run graphify --help to confirm. The installation docs cover Windows PowerShell and other edge cases.

Register Graphify with your assistantLink to Register Graphify with your assistant

Installing the package gives you the CLI. To let your assistant use it, register the skill:

bash
graphify install

On its own, graphify install targets Claude Code. For other assistants, name the platform:

bash
graphify install --platform codex
graphify install --platform cursor
graphify antigravity install

Add --project to write the skill into the current repository instead of your user profile, so teammates get it from git. For Claude Code, graphify claude install also writes a section to CLAUDE.md and registers hooks, so the assistant checks the graph before searching files.

Build your first graphLink to Build your first graph

Open your assistant in a repository and send:

text
/graphify .

The graph lands in graphify-out/. To build a code-only graph from the terminal, with no model calls at all, run:

bash
graphify extract . --code-only

A code-only build parses files with tree-sitter on your machine and makes no model calls. Build time and graph size depend on the codebase. Then ask the graph a question:

bash
graphify query "how does authentication work"
graphify path "Client" "Transport"

The open-source CLI supports about 40 languages. Our comparison of codebase context tools shows where a graph fits.

Should I use hosted Graphify instead?Link to Should I use hosted Graphify instead?

Use hosted Graphify when your code is on GitHub or GitLab and you want graphs rebuilt on every push, shared with a team and used in pull request reviews. Reviews and formal verification run on GitHub repositories. You install nothing locally: connect a repository at app.graphify.com, add the hosted MCP server and sign in with OAuth.

For Claude Code, the whole setup is one command:

bash
claude mcp add --transport http --scope user graphify https://api.graphify.com/mcp

Then run /mcp in Claude Code and choose Authenticate. You can start on the Free plan. See the integrations overview for other clients.

How do I upgrade or uninstall Graphify?Link to How do I upgrade or uninstall Graphify?

Upgrade the CLI with uv tool upgrade graphifyy, or pipx upgrade graphifyy if you used pipx. Uninstall with uv tool uninstall graphifyy. Graph files live in each repository's graphify-out/ folder, so delete that folder to remove a graph. Hosted Graphify updates itself, and you remove repositories from the console.

After an upgrade, run graphify install again so the assistant skill matches the new version.

Try itLink to Try it

The fastest start is hosted: sign in at app.graphify.com and connect a repository. For a local graph:

bash
uv tool install graphifyy
graphify install

New to the idea? Start with what is a code knowledge graph.

Related reading: giving Claude Code a code knowledge graph and change impact analysis. Sources: pipx docs, graphifyy on PyPI.

FAQLink to FAQ

Why is the package graphifyy and not graphify?Link to Why is the package graphifyy and not graphify?

The PyPI name has two y's. The command is still graphify. Other similarly named packages are not from Graphify Labs.

Does the Graphify CLI need an account or API key?Link to Does the Graphify CLI need an account or API key?

No. It is free under Apache 2.0, needs no account, and code-only builds make no model calls.

Which Python version does Graphify need?Link to Which Python version does Graphify need?

Python 3.10 or newer.

Can I use uvx instead of installing?Link to Can I use uvx instead of installing?

You can run a one-off command with uvx, but a persistent install with uv tool install graphifyy is better for assistants and hooks.

Does hosted Graphify need the CLI?Link to Does hosted Graphify need the CLI?

No. Hosted Graphify builds the graph for you. Your assistant connects to https://api.graphify.com/mcp.

Try Graphify now.

Try for free