Docs · Install

Install Graphify.

Two steps: install the CLI from PyPI, then register the /graphify skill with your assistant. No account, no API key, nothing leaves your machine.

Install the package

Graphify ships as a single Python package on PyPI. Heads up on the spelling: the package is graphifyy (double y), the command it installs is graphify. Python 3.10 or newer is required. uv tool install and pipx isolate the package in their own environment, which avoids the interpreter-mismatch problems plain pip can hit on Mac and Windows.

terminal

# recommended: isolated env via uv$ uv tool install graphifyy✓ installed graphify # alternatives:# pipx install graphifyy# pip install graphifyy     (may need PATH setup, see troubleshooting)

Register the skill

Installing the package gives you the binary; it does not touch your assistant yet. graphify install copies the /graphifyskill into your assistant's config. Bare graphify install targets Claude Code by design (Windows is auto-detected); name any other assistant explicitly.

terminal

# Claude Code (the default platform):$ graphify install✓ /graphify skill registered # any other assistant, by name:$ graphify cursor install$ graphify codex install$ graphify gemini install # same thing via a flag:$ graphify install --platform codex

Per-assistant commands

Every supported assistant has a dedicated subcommand (each also accepts uninstall and --project). One quirk worth knowing: Codex invokes the skill as $graphify instead of /graphify, and PowerShell users should type graphify . without the leading slash.

terminal

graphify claude install        # Claude Codegraphify cursor install        # Cursorgraphify codex install         # Codex   (invoke with $graphify)graphify gemini install        # Gemini CLIgraphify copilot install       # GitHub Copilotgraphify vscode install        # VS Codegraphify aider install         # Aidergraphify opencode install      # OpenCodegraphify amp install           # Ampgraphify devin install         # Devingraphify kilo install          # Kilographify kiro install          # Kirographify droid install         # Factory Droidgraphify trae install          # Trae   (trae-cn for the CN build)graphify claw install          # OpenClawgraphify pi install            # Pigraphify hermes install        # Hermesgraphify codebuddy install     # CodeBuddygraphify antigravity install   # Antigravitygraphify agents install        # generic .agents/skills (alias: skills)

Per-assistant setup guides, with what each install actually writes, live under integrations.

Scopes & strict mode

By default the skill installs into your user config, so it works in every repo. --project installs into the current repo instead. On Claude Code, --strictupgrades the default soft nudge ("query the graph before grepping") into a hook that blocks the first raw source read of a session and redirects it to the graph, once per session.

terminal

# project scope: install into the current repo instead of your user config$ graphify install --project # strict mode (Claude Code): block the first raw source read of a session# and redirect it to the graph, then revert to a soft nudge$ graphify install --project --strict# or: graphify claude install --strict# runtime toggle: GRAPHIFY_HOOK_STRICT=1 (on) / 0 (off)

Verify

Confirm the CLI is on your PATH, then build your first graph from inside the assistant. The tutorial walks the whole first run step by step.

terminal

$ graphify --versiongraphify 0.9.22 # then, inside your assistant, build your first graph:/graphify .

Troubleshooting

graphify: command not found right after installing? uv and pipx put commands in ~/.local/bin, which is not on PATH on a fresh macOS + zsh setup:

terminal

# uv installs put the command in ~/.local/bin; add it to PATH:$ uv tool update-shell# (pipx equivalent: pipx ensurepath)# then open a new terminal # plain pip: add the bin dir to PATH yourself, or skip PATH entirely:$ python -m graphify --version

Running via uvx / uv tool run without installing? Name the package, not the command:

terminal

# uvx reads the first word as a PACKAGE, and the package is graphifyy:$ uvx --from graphifyy graphify install # plain "uvx graphify" fails: there is no PyPI package named graphify

Seeing ModuleNotFoundError: No module named 'graphify' after a plain pip install on Mac or Windows? The skill resolves Python at runtime, and plain pip can point it at a different environment than the one holding the package. Reinstall with uv tool install or pipx, which avoid this entirely.

Uninstall

One command removes the skill from every assistant it was installed into; the graph files stay unless you ask for --purge.

terminal

# remove the skill from every detected assistant:$ graphify uninstall # also delete the graphify-out/ directory:$ graphify uninstall --purge # one assistant only:$ graphify claude uninstall

Next: the tutorial (first graph, first query), the CLI reference, or the README on GitHub for platform-specific notes.