Graphify GitHub (2026): Install the Code Graph Skill
By Antigravity Skills editorial team · Last updated
Short answer: Graphify GitHub means Graphify-Labs/graphify, an Apache-2.0 open-source tool that turns a codebase, plus its docs and PDFs, into a knowledge graph you can query. Install it with uv tool install graphifyy, run graphify install, then type /graphify . in your coding agent. It ships an installable agent skill for Claude Code, Codex, Cursor and Antigravity.
Skill page: the agent skill has its own guide, Graphify agent skill: install and use, with the install commands and the commands the skill gives your agent.
Unofficial guide. Antigravity Skills is not affiliated with Graphify Labs. The official sources are the repository github.com/Graphify-Labs/graphify and the docs at docs.graphify.com. Facts below were checked on November 19, 2026 against the GitHub API, the README, the repository tree and the skills.sh listing. The project releases almost daily, so confirm commands at the source.
At a glance
| What it is | A CLI plus an agent skill: it maps code, docs, PDFs, images and video into a knowledge graph you query instead of grepping |
| Repository | Graphify-Labs/graphify; the old path safishamsi/graphify redirects to it |
| Stars and forks | 150,310 stars and 14,454 forks on November 19, 2026 |
| Created | April 3, 2026 |
| Last push | November 17, 2026 (21:21 UTC); latest release v0.9.112 on November 17, 2026 |
| License | Apache-2.0 per the GitHub API and pyproject.toml; the repository also keeps a LICENSE-MIT file |
| Package | PyPI graphifyy (double y); the command is graphify; Python 3.11 or newer |
| Cost | Free to run. Code is parsed locally; docs, PDFs and images use your assistant's model or an API key you configure |
| Hosted version | Graphify Cloud, a separate paid service, described in the README |
| Agent skill | Yes. The skill is graphify/skill.md in the package, with per-platform variants, and graphify install writes it as SKILL.md |
Which Graphify GitHub repository is the right one?
Several GitHub repositories use the name or build on it. These are the ones we saw in a GitHub search on November 19, 2026:
| Repository | What it is | Stars |
|---|---|---|
| Graphify-Labs/graphify | The knowledge-graph tool and /graphify skill this page covers; homepage is graphify.com | 150,310 |
| lucasrosati/claude-code-memory-setup | A setup that combines Obsidian with Graphify for Claude Code memory; not Graphify itself | 1,073 |
| kbastani/graphify | A separate older project: a Neo4j unmanaged extension for text classification | 453 |
| crabbuild/compass | A Rust knowledge-graph engine that says it is inspired by Graphify | 176 |
| elbruno/graphify-dotnet | A .NET 10 port of graphify | 99 |
| raufer/graphify | A separate library for parsing unstructured text into graphs | 99 |
Two details prevent wrong installs. First, the README says the official PyPI package is graphifyy with two y, and that other graphify* packages on PyPI are not affiliated. Second, some third-party pages and the graphify.com FAQ still point at safishamsi/graphify. When we called the GitHub API with that path it returned Graphify-Labs/graphify, so both lead to the same repository.
What Graphify does
You type /graphify in your AI coding assistant and it builds a graph of the project. The README lists these parts:
- Local code parsing. Code is parsed with tree-sitter, about 40 languages, with no LLM and nothing leaving your machine.
- Labelled edges. Every connection is tagged
EXTRACTED(explicit in the source) orINFERRED(resolved by Graphify), so you can see what was read and what was guessed. - Not a vector index. There are no embeddings; you traverse a real graph with
query,pathandexplain. - God nodes and communities. It ranks the most connected concepts and splits the graph into subsystems.
- Beyond code. Docs, PDFs, images and video or audio go into the same graph through an optional semantic pass.
A run writes three files into graphify-out/:
| File | What it is for |
|---|---|
graph.html | Open in a browser to click, filter and search the graph |
GRAPH_REPORT.md | The highlights: key concepts, surprising connections, suggested questions |
graph.json | The full graph, so you can query it later without re-reading your files |
Requirements
- Python 3.11 or newer.
pyproject.tomlsays>=3.11,<3.15. - uv or pipx. The README recommends
uv;pipxand plainpipalso work. - An AI coding assistant, if you want the
/graphifycommand. You can also use the CLI alone. - No API key for a code-only graph. The README says
graphify extractruns offline on code. Mixed repos with docs, PDFs or images need your assistant's model, or one of the supported keys or a local Ollama.
Install Graphify GitHub's CLI and skill
These are the README's two commands:
uv tool install graphifyy # or: pipx install graphifyy
graphify install # detects your assistant and registers the skillThen open your assistant in the project and run:
/graphify .If graphify is not found after installing with uv, the README says to run uv tool update-shell.
Install in Antigravity, Claude Code and Codex
The README gives one command per platform. These three matter most for this site:
| Agent | Command | Where the skill goes |
|---|---|---|
| Claude Code | graphify install, or graphify claude install | .claude/skills/graphify/SKILL.md for a project install, per the README |
| Google Antigravity | graphify antigravity install | In the repository's own test, a project install writes .agents/skills/graphify/SKILL.md plus .agents/rules/graphify.md and .agents/workflows/graphify.md |
| Codex | graphify install --platform codex | Project installs write under .claude/skills/ or .agents/skills/ per the README's examples; Codex uses $graphify, not /graphify |
Add --project to any of them to write into the current repository instead of your user profile, for example graphify install --project --platform codex. The README lists more than 20 platforms, including Cursor, Gemini CLI, GitHub Copilot CLI, OpenCode, Kiro and Hermes. For the general skills folders, see how to install skills in Antigravity, Claude Code and Codex.
There is also a listing on skills.sh: npx skills add graphify-labs/graphify. On November 19, 2026 it showed 14.6K installs for one skill, graphify. The README's own route is graphify install, which also writes the always-on hook files, so we treat it as the main one.
How to use it
The commands in the README, run inside your agent:
/graphify . # build the graph for this folder
/graphify ./docs --update # re-extract only changed files
/graphify query "what connects auth to the database?"
/graphify path "UserService" "DatabasePool"
/graphify explain "RateLimiter"
/graphify add https://arxiv.org/abs/1706.03762 # fetch a paper into the graphTo make the assistant consult the graph before it greps, run graphify claude install (or the matching platform command) once in the project. On Claude Code and Gemini CLI it adds a hook that nudges the agent toward the graph; on Codex, OpenCode and Cursor it writes instruction files such as AGENTS.md. Run graphify hook install and the graph rebuilds on commit and branch checkout. To remove it, graphify uninstall.
Limits and privacy
- Semantic extraction is not free of cost. Code is local, but docs, PDFs and images go to the model your assistant uses, or to the provider whose key you set. For data-residency needs the README points to
GRAPHIFY_BACKEND=ollama. - Inferred edges are guesses. Check the
INFERREDtag before you trust a link between two files. - Query logging. The README says every
query,pathandexplaincall is logged to~/.cache/graphify-queries.log, and you can turn it off withGRAPHIFY_QUERY_LOG_DISABLE=1. It also says there is no telemetry. - Fast-moving project. The release number changed from v0.9.15 in one third-party page to v0.9.112 on the repository, so numbers in articles go stale quickly.
- Benchmarks are the maintainers'. The README's recall figures come from its own harness. We did not reproduce them.
- We did not run it. This guide comes from the README, the repository files and the GitHub API. We did not install the package or build a graph ourselves.
FAQ
What is the Graphify GitHub repository?
It is Graphify-Labs/graphify, an open-source tool that builds a queryable knowledge graph from code and other files. It was created on April 3, 2026 and had 150,310 stars when we checked.
Is Graphify free and open source?
The GitHub API and pyproject.toml list Apache-2.0, and the tool runs locally with no cost beyond the model you choose for docs and images. Graphify Cloud is a separate paid hosted product.
How do I install Graphify?
Run uv tool install graphifyy, then graphify install, then /graphify . in your assistant. The package name has two y letters.
Does Graphify have an agent skill?
Yes. The package contains graphify/skill.md with the name graphify in its frontmatter and per-platform variants, and graphify install writes it as SKILL.md. It is also listed on skills.sh.
Does Graphify work with Antigravity?
The README lists graphify antigravity install, and the repository has a test for it that writes the skill, a rules file and a workflow file under .agents/.
Does Graphify send my code to a server?
Code files are parsed locally with tree-sitter. Docs, PDFs and images are sent to your assistant's model or to a provider you configure, according to the README's privacy section.
Why is the PyPI package called graphifyy?
The README says graphifyy with two y is the official package and that other graphify* packages are not affiliated. The command you run is still graphify.
Related guides
- Graphify agent skill: install and use
- How to install skills in Antigravity, Claude Code and Codex
- Find skills in Antigravity
- Agent Skills GitHub: official repos and how to use them
- claude-mem: persistent memory for Claude Code
- Browse all skills
Sources
- Graphify-Labs/graphify on GitHub: README,
LICENSE,pyproject.toml,graphify/skill.md,tests/test_antigravity_install.py, GitHub API, checked November 19, 2026 - docs.graphify.com, linked from the README
- skills.sh listing for graphify-labs/graphify