hatz-provider

Adds Hatz AI as a model provider to your coding agents. One command.

npx hatz-provider install

Built by Sam King

What it does

Points your coding agents at the Hatz AI API — 80+ models (OpenAI, Anthropic, Google, xAI, DeepSeek, Meta, Mistral, and more) through one API key.

Each agent gets the API surface it expects:

• omp, pi, Claude Code → Anthropic Messages (/v1/anthropic)
• Hermes, OpenClaw → OpenAI Completions (/v1)

Config blocks are guarded with comment markers. Reinstalling or updating only touches the Hatz section and leaves your other providers alone.

Quick start

npx hatz-provider install

You'll be prompted for your API key (get one → Workspace → API Keys), then Hatz is installed into every supported agent detected on your machine.

Want just one agent?

npx hatz-provider install omp
npx hatz-provider install hermes

How it picks agents

install and uninstall with no agent name auto-detect which supported agents are set up on your machine — by checking for ~/.omp, ~/.pi, ~/.hermes, ~/.claude, ~/.openclaw — and only touch those. Naming an agent explicitly (e.g. install omp) always installs it, detected or not.

Commands

install [agent]     Add Hatz provider (fetches live model catalog)
update [agent]      Refresh models from the catalog
uninstall [agent]   Remove Hatz provider
status              Show which agents have Hatz installed
list                Print all available models
usage               Show Hatz credit usage
catalog             Dump omp models.yml block
-n, --dry-run       Preview changes without writing files

Setting the key instead of a prompt

For scripts or CI, set the key in your shell before running:

# PowerShell
$env:HATZ_API_KEY = "hzat-..."

rem cmd.exe
set HATZ_API_KEY=hzat-...

# bash / zsh
export HATZ_API_KEY="hzat-..."

How it works

hatz-provider fetches the Hatz model catalog when you install and writes the right config for each agent. Where the agent can resolve env vars (omp, pi, Hermes via key_env), the config references HATZ_API_KEY. Claude Code stores the key in its settings.json env block. The key is never sent anywhere except to the Hatz API at runtime.

Run install or update again to refresh the model list. Only the Hatz block is changed.

Requirements

• Bun installed globally (the CLI runs on Bun)
• A Hatz AI API key

License

MIT