Roma: the to-do list for your AI agents
We're living in exciting times. You can write code with Codex. Create animations with Opus. Muse has your credit card. Grok runs your X account. New agents are popping up while the ones we already use keep getting better and better.
But working across them has become a mess. We tell each agent the same thing over and over, copy and paste between chats, and chase down work we've already done. We're tired of babysitting our agents.
How we got here
We launched Roma four weeks ago as the to-do list that does itself. Immediately, people started connecting their agents, which was exciting to see. Among our most active users, many are using the product more than 3 hours a day, some with upwards of 100,000 tool calls per user.
They track which agent is waiting on a decision. They review work in Roma and pass it on to another agent. They have their project plan in Roma so that when a new chat with an agent starts, they can pick up where the last agent left off.
They weren't just keeping a to-do list for themselves. They were sharing it with their agents, and it became a better way to manage the work together.
This was when we doubled down and put in the hard work to make sure the work could truly flow in both directions. Connecting agents to read and write tasks was only the start. We wanted you to be able to hand an agent work, see what it was doing, and answer when it needed you. And for another agent to pick up where it left off, with a record of what happened along the way.
Introducing Roma 2.0
Today we're launching Roma 2.0, the world's first to-do list for your AI agents.
Roma is the single source of truth for you, your team, and all of your AI agents. A single place where you can work together, share context, and delegate to a person or agent inside any task.
It's designed for agents to work in, but elegant and simple enough for people to read. The context is shared and stored, so new teammates or agents can catch up from the task's history.
What that looks like
Hand any task to any agent
In Roma, the work moves both ways. Give a task to Claude, Codex, Muse, or Roma's own Operator, and it comes back with the result or a question for you. Your agents can hand work to each other too. You can be in ChatGPT and send a task to Muse without leaving the chat. Whoever picks it up next starts from the task, never from scratch.
Everything is on the record
Every task keeps a full history: each decision, edit, change, and who made it. Whether that was you, a teammate, or an agent. Roma summarizes it so you can catch up quickly, and the full record is always there when you want the details. It's written for humans to read and agents to work from.
Lists for everything and everyone
Make a list for anything: customers, content, a product roadmap. Share it with your partner or team, and bring everyone's agents into the work. Your Claude and their ChatGPT can work on the same task. Nobody has to agree on one agent.
Get started
This is how we've wanted to work with our agents: together, from one list, without starting over every time we switch.
We built Roma 2.0 for that, and we can't wait to see what you and your agents do with it. Connect your agents at roma.app/agents and give them something you've been meaning to get done.