AI agents you message like people
Each agent has one job, such as paying your bills or looking into a question, and a name that says what it is. Write to one, and it replies and gets to work. Each keeps notes in files of its own, so it remembers what you told it from one conversation to the next.
A Billion Agents is in private beta. Join the waitlist, and we'll email you when there's room.
What your agents do
A conversation for every task
Start a new conversation for each thing you want done, even with an agent you've talked to before. Every conversation sits in one list, newest first, and you can pick up any one again. Tag another agent with @ and its name, and it answers once in that conversation. Attach photos and PDFs to a message, and the agent reads them with it. Replies appear as the agent writes them, and Stop ends one you no longer need.
They remember what you tell them
Tell the agent that pays your bills the rent is $2,150 on the 1st, and it files a note in knowledge/payables/recurring.md, one of the files in a workspace of its own. It reads those files again when they bear on what you ask, in any conversation. Agent info shows what an agent is for, what wakes it and each knowledge file it keeps.
They work on a computer, in a browser and on GitHub
An agent can start a computer of its own, run commands on it and copy files between it and its workspace. It attaches the files it makes to its replies, and draws the charts and pictures you ask for with code. The computer pauses when it sits idle and is discarded when the agent's working session ends. In a real browser, an agent reads web pages and carries out tasks on sites. When a site wants a sign-in, the agent hands you the browser to sign in yourself. Before a step that's hard to undo, such as placing an order, it asks, and acts only if you allow it. Connect a GitHub account, and your agents can work with its repositories, issues and pull requests, wherever you install the GitHub App.
Describe the agent you need
You start with one agent, which answers your questions about the app from the user guide. To make another, describe its job in a message, such as "Keep track of my invoices." It drafts itself as a card with a name, a purpose and the files it starts with. Reply to change anything, its name included, and Create makes it an agent. Ask one of your agents to set up another for a project, and it creates a delegate: an agent of its own that works on the project and reports back in the conversation you asked in.
See how an agent works
Agents run on Claude, Anthropic's model. The inspector shows an agent's instructions, every file in its workspace, each piece of work it did and how long it took, and what it's doing right now. You don't need it to use the app; it's there when you want to check an agent's work.
Joining the private beta
The app runs in a web browser, on a laptop or a phone. Agents work when you message them, when one agent hands another work, and at times you ask for, such as every weekday at 7:30. They can't yet receive email. Pricing has not been decided. People join by invitation, from this waitlist.