Skill guide · checked November 19, 2026

Graphify Agent Skill: Install and Use

Short answer: Graphify is an agent skill that maps a codebase and its docs into a knowledge graph your agent queries instead of grepping. Install the CLI with uv tool install graphifyy, run graphify install, then type /graphify . in your agent. The Graphify-Labs/graphify repository ships the skill for Claude Code, Codex, Antigravity and 20 more platforms.

Want the tool itself first? Read Graphify GitHub: Install the Code Knowledge Graph Skill.

Install the Graphify skills

The skill is installed by the graphify command, not copied by hand: the package contains skill.md and per-platform variants, and graphify install writes the right one as SKILL.md for your agent. The PyPI package name has two y letters. The skill is only instructions plus the CLI it calls, so the CLI must be on your PATH.

# install the CLI (needs Python 3.11+), then register the skill
uv tool install graphifyy      # or: pipx install graphifyy
graphify install               # Claude Code, detects your assistant

# other agents (README)
graphify antigravity install   # Google Antigravity
graphify install --platform codex
graphify install --project     # write into this repo instead of your profile

# skills.sh listing (14.6K installs when we checked)
npx skills add graphify-labs/graphify

Source: Graphify-Labs/graphify and the official docs. We do not host a copy, so you always install the current version.

What each Graphify skill does

The package ships one skill, graphify, with a usage list in its skill.md. This table shows the commands in our words; read the upstream file for the exact text.

SkillWhat it doesUse when
Build the graph/graphify . maps the current folder (or a path, or a GitHub URL it clones) and writes graph.html, GRAPH_REPORT.md and graph.json into graphify-out/.You open an unfamiliar repository and want the structure first.
Query, path and explainquery asks a plain-language question and returns a scoped subgraph; path traces how two things connect; explain describes one concept with its edges.You would otherwise grep through many files to answer a question.
Update incrementally--update re-extracts only new or changed files; --cluster-only reruns community detection on the existing graph; --force allows a full rebuild.The code changed since the last run.
Add outside material/graphify add <url> fetches a paper, page or video into ./raw and updates the graph, with optional author and contributor tags.Docs or papers belong in the same graph as the code.
Export and serveFlags export an SVG, GraphML, a Neo4j or FalkorDB Cypher file, an Obsidian vault or a markdown wiki, and --mcp starts an MCP stdio server.Another tool or another agent needs the graph.
Always-on guidancegraphify claude install (or the matching platform command) adds a hook or instruction file so the agent checks the graph before searching raw files.You want the agent to use the graph without being asked.

Map a repository with Graphify and an agent

Based on the README and skill.md. Code is parsed locally; docs, PDFs and images use your assistant model or a key you configure. We did not run the pipeline ourselves.

  1. Install the CLI with uv tool install graphifyy and register the skill with graphify install (or the command for your agent).
  2. Open your agent in the project folder and run /graphify . ($graphify . in Codex).
  3. Open graphify-out/graph.html and read GRAPH_REPORT.md for the main concepts and suggested questions.
  4. Ask questions with /graphify query "<question>" and trace links with /graphify path "A" "B".
  5. Run /graphify . --update after changes, or graphify hook install to rebuild on commit.

License and what we did not verify

The GitHub API and pyproject.toml list Apache-2.0, and the repository also keeps a LICENSE-MIT file. We link to the skill instead of hosting a copy. Graphify Cloud, a hosted paid product described in the README, is separate from the open-source tool. The README says the code path sends nothing off your machine; we did not verify that ourselves. Of the 20 or so platforms, we checked only the install commands, not the results. See also how to install skills in Antigravity, Claude Code and Codex and our full Skills library.

FAQ

Does Graphify have an agent skill?

Yes. The Graphify-Labs/graphify package contains skill.md named graphify plus per-platform variants, and graphify install writes it as SKILL.md. It is also listed on skills.sh.

How do I install the graphify skill?

Run uv tool install graphifyy, then graphify install. For Antigravity use graphify antigravity install, and for Codex graphify install --platform codex. Add --project to install into the current repository.

Is the skill the same as the pip package?

No. The package is the CLI that builds and queries the graph. The skill only teaches your agent the /graphify commands, so you need both, and graphify install registers the skill for you.

Does Graphify need an API key?

Not for a code-only graph, according to the README. Docs, PDFs and images need your assistant model, a supported provider key or a local Ollama.

Does Antigravity Skills host a copy of the skill?

No. This page links to the official repository and does not redistribute its files, so you install the current version.