CLI

Requirements

• Node.js >= 24, or
• Bun >= 1.1

Either runtime works everywhere below — pick whichever you already have installed.

Install (recommended)

One-liner installers that install Bun if needed, then install the latest hoomanjs globally:

curl

curl -fsSL https://raw.githubusercontent.com/vaibhavpandeyvpz/hooman/main/install.sh | bash

PowerShell

irm https://raw.githubusercontent.com/vaibhavpandeyvpz/hooman/main/install.ps1 | iex

npm

npm i -g hoomanjs

bun

bun add -g hoomanjs

Then run:

hooman

Run without installing

The fastest way to try Hooman without a global install:

npx hoomanjs

Or with Bun:

bunx hoomanjs

Both download hoomanjs on first use and cache it for next time — no separate install step, and nothing left behind if you don’t keep using it.

First run

1. Run hooman (or npx hoomanjs). If ~/.hooman/config.json is missing, Hooman opens a setup wizard first:
   • pick an inference provider and enter credentials (validated against the provider)
   • pick a search provider (DuckDuckGo needs no key; others are validated with a test search)
   • write config.json with the available chat LLMs for that provider
2. Chat starts automatically after setup. Re-run setup anytime with hooman setup.
3. Use /config inside chat (or hooman config) for advanced settings — MCP, skills, prompts, tools, and more.
4. Use hooman exec "your prompt" for one-off tasks once you’re set up.

The setup wizard defaults to local llama.cpp (Gemma 4 E2B / Qwen3.5 2B on Q4_K_M) so you can start without an API key; on Apple Silicon you can choose MLX instead. Hosted providers (Anthropic, OpenAI, Google, …) are available in the same flow. The first local model you use downloads from the Hugging Face Hub on demand with a live progress bar. See Models for the full list.

Run from source

Prefer to build from the repository instead of the published package:

git clone https://github.com/vaibhavpandeyvpz/hooman.git
cd hooman
npm install
npm run dev -- --help

Or build and link the CLI locally:

npm run build
npm link
hooman --help

Next steps

• CLI reference — every command, flag, and in-chat slash command.
• Configuration — ~/.hooman layout and repo-local overlays.
• Models — connect Anthropic, OpenAI, Google, Ollama, and more.
• VS Code quickstart — prefer a chat panel in your editor instead?