What is the Graphify skill, and how do you use it?
· 6 minute read
The Graphify skill lets your coding agent build and query a local knowledge graph of the repository. You install the open-source package once, register the skill, then type a short command in the agent. The agent builds the graph and reads GRAPH_REPORT.md before it answers questions about your code.
This is the step by step guide for Claude Code, Codex and Cursor. The Graphify skill quickstart is the command reference this post follows. If you are choosing between the local skill and a hosted setup in Claude Code, start with Graphify in Claude Code: hosted MCP or a skill?. That post is the comparison. This one is how to run the skill.
How do you install the Graphify skill?Link to How do you install the Graphify skill?
You need Python 3.11.5 or later, and a coding agent that graphify install supports. Claude Code is the default.
Install the package with uv, which is the recommended path. It can install a suitable Python for you and keeps Graphify in its own environment:
uv tool install graphifyyThe package name is graphifyy, with two y's. That install gives you the graphify command. If you already manage command line tools with pipx, pipx install graphifyy works the same way. If you manage Python yourself, pip install graphifyy is fine too.
Then register the skill:
graphify installWith no flag, this targets Claude Code. It copies the skill to ~/.claude/skills/graphify/SKILL.md and registers it in ~/.claude/CLAUDE.md. For Cursor or Codex, pass --platform:
graphify cursor install
graphify install --platform codexIf the agent does not pick up the skill, restart it so it reloads skills. For Claude Code, check that ~/.claude/skills/graphify/SKILL.md exists. For the other agents, run the matching graphify install --platform command again.
How do you use Graphify in Claude Code, Cursor and Codex?Link to How do you use Graphify in Claude Code, Cursor and Codex?
Open the agent in the repository you want to map.
In Claude Code and Cursor, select the Graphify skill and enter:
/graphify .In Codex, select the Graphify skill from the $ picker and enter:
$graphify map the current repositoryThe agent runs detection, extraction and clustering, then summarizes the report. Code is parsed locally. Documents, papers and images go through the agent's own model.
Once the graph is built, ask a structural question, such as how authentication reaches the database, and tell the agent to use Graphify. You can also query from a terminal:
graphify query "how does authentication reach the database"The skill walks the graph and returns the nodes and edges that answer the question. To confirm the graph is there, run graphify god-nodes --top 5 from the folder that contains graphify-out/. You should see the five most connected symbols in the code.
What lands in graphify-out/?Link to What lands in graphify-out/?
After the first run you will find a graphify-out/ folder in the repository:
graphify-out/
├── graph.json
├── graph.html
├── GRAPH_REPORT.md
├── manifest.json
└── cache/graph.html is the visual map. Open it in a browser to click through nodes, filter and search. GRAPH_REPORT.md is the written report the agent reads, with key concepts and suggested questions. graph.json is the full graph, so you can query it later without reading the repository again.
Commit graph.json and GRAPH_REPORT.md if you want teammates and CI to share the same graph. Add cache/ to .gitignore.
How do you keep the graph up to date?Link to How do you keep the graph up to date?
From a terminal, refresh the code graph with:
graphify update .That command re-extracts only new or changed code. It does not re-read documents, papers or images. When those change, run this in your agent:
/graphify . --updateYou can also install git hooks so the code graph refreshes in the background:
graphify hook installThe hooks run after commits, branch switches, merges, pulls and rebases, so you do not need to run graphify update . after git pull. Check they are in place with graphify hook status. A query right after a commit or pull can briefly see the previous graph, because the refresh runs in the background. The incremental updates page covers what each refresh includes.
Where does hosted Graphify fit?Link to Where does hosted Graphify fit?
The skill keeps the graph on your machine, in graphify-out/. That is the right path when the code lives only on your laptop, or you want to stay offline.
Hosted Graphify is shared codebase context for a team and the agents that work with it: a code knowledge graph, a Wiki, Memory and pull request reviews. Agents connect to the hosted MCP server at https://api.graphify.com/mcp and sign in with OAuth. There is no API key to paste. A push to the default branch starts a new build of that commit. Your agents keep using the previous graph until the new one is ready.
For a fuller comparison of the two Claude Code setups, use the hosted MCP or skill guide. Open-source Graphify and hosted Graphify covers the same split across the product.
Try itLink to Try it
Connect a repository at app.graphify.com and point your agent at https://api.graphify.com/mcp. Or build the graph on your machine:
uv tool install graphifyy
graphify installThen, in Claude Code or Cursor, type /graphify .. In Codex, select the skill and enter $graphify map the current repository.
Open-source Graphify is licensed under Apache 2.0. The package lives at Graphify-Labs/graphify. Related reading: what a code knowledge graph is, Graphify in Claude Code: hosted MCP or a skill? and how to use Graphify with coding agents. Sources: the Graphify skill quickstart, incremental updates, how Graphify works and the Claude Code, Cursor and Codex integration pages.
FAQLink to FAQ
What is the Graphify skill?Link to What is the Graphify skill?
The Graphify skill lets your coding agent build and query a local knowledge graph of the repository. You install graphifyy once, run graphify install, and type a short command in the agent so it can build the graph and read the report before it answers questions about your code.
How do you use Graphify in Claude Code, Cursor and Codex?Link to How do you use Graphify in Claude Code, Cursor and Codex?
Install graphifyy, then run graphify install for Claude Code, graphify cursor install for Cursor, or graphify install --platform codex for Codex. In Claude Code and Cursor, type /graphify . In Codex, select the skill and enter $graphify map the current repository.
What files does the Graphify skill write?Link to What files does the Graphify skill write?
It writes a graphify-out folder that includes graph.html, GRAPH_REPORT.md and graph.json. Open graph.html in a browser to explore the graph. Commit graph.json and GRAPH_REPORT.md if teammates and CI should share the same graph, and keep the cache folder out of git.
How do you update the graph after the code changes?Link to How do you update the graph after the code changes?
Run graphify update . in a terminal to refresh the code graph. For document, paper or image changes, run /graphify . --update in the agent. You can also install git hooks so the graph refreshes after commits, branch switches, merges, pulls and rebases.
When should I use hosted Graphify instead of the skill?Link to When should I use hosted Graphify instead of the skill?
Use hosted Graphify when you want shared codebase context for a team: a code graph, a Wiki, Memory and pull request reviews, served to agents at api.graphify.com/mcp. The skill keeps the graph on your machine in graphify-out.
What is Graphify AI?Link to What is Graphify AI?
Graphify is shared codebase context for coding agents and the people who work with them: a code graph, a Wiki, Memory and pull request reviews. The Graphify skill is the local way to build and query that graph from your coding agent.
Do I need an API key to use the Graphify skill?Link to Do I need an API key to use the Graphify skill?
No. The skill runs on your machine. Hosted Graphify signs in with OAuth in the browser, so there is no key to paste there either.