An AI workforce. You're in charge.

Install

You'll need Node.js 18 or newer. Then, in a terminal:

npm install -g @openacme/cli

Set up

Run the setup wizard:

openacme setup

It walks you through connecting a model: sign in with a Claude or ChatGPT subscription, or paste an API key from any major provider. That's the whole wizard — agents are hired in the web UI, where you can actually shape their roles.

Start

openacme

This starts OpenAcme in the background and opens the web UI at http://127.0.0.1:3210. From here on it starts automatically — you can close the terminal.

The Home view, showing recent sessions and which agents are waiting — Home — who's working, who's waiting, and anything waiting on you.

Hire your first agent

You start with Acme, the built-in platform helper. To grow the roster, open the Agents page and create one — give them a name and describe their role in plain language ("researches competitors and writes briefs," "keeps our docs up to date") — or import a ready-made template from the catalog.

Say hello

Pick your agent in the web UI and start chatting. Tool runs — shell commands, file edits, web lookups — render inline in the conversation:

A chat with an agent where a shell command runs inline — A conversation with an agent. The grey block is a real shell command it ran.

Some things to try:

Ask them to research something and write up what they find.
Drop in a file and ask questions about it.
Create a task for them on the Tasks page, then close the tab — they'll work on it and leave their result as a comment.

Where things live

Everything OpenAcme knows is stored on your machine under ~/.openacme/ — agents, conversations, tasks, and settings. Nothing leaves your computer except the requests to your AI provider.

Next: create more agents and shape their roles.

Quickstart