ax talks to your team's Axonate gateway — every model (real provider APIs + free local models)
behind a single OpenAI-compatible endpoint. Built for small teams who need AI without a
plan-per-person bill: one gateway, one key, per-user budgets.
macOS (Homebrew)
brew install Axonate/tap/axpip — macOS / Linux / Windows (zero dependencies, any Python 3.8+)
pip install axonate-ax # or: pipx install axonate-axWindows (PowerShell or CMD)
py -m pip install axonate-ax
# 'ax' is now on PATH; if not, restart the terminal or add the Scripts dir to PATHWithout installing
python -m axonate_cli "hello"mkdir -p ~/.config/axonate
cat > ~/.config/axonate/config <<EOF
AXONATE_URL=https://axonate.yourdomain.com # or http://127.0.0.1:4100 locally
AXONATE_KEY=sk-your-virtual-key
EOFOr use env vars AXONATE_URL / AXONATE_KEY.
ax "explain this error" # smart auto-routing picks the model
ax -m claude "write a haiku" # force a model
git diff | ax "review this" # pipe stdin
ax -s work "remember X" # session continuity
ax -s work "what did I say?" # ...same session recalls it
ax -s work --new "fresh start" # reset a session
ax --no-stream "..." # disable streamingThe gateway prints which model handled it ([route: codex]) so you always know where your
tokens went.
| Without | With Axonate |
|---|---|
| One AI plan per person | One shared gateway, every model, per-user budgets |
| No idea who spent what | Per-user budgets + spend tracking |
| Pick the model manually | auto routes cheap tasks to cheap models |
Set up the gateway: Axonate/axonate · Docs: Axonate/docs
MIT © CloudDrove