Put OpenAI, Claude, Cursor, Antigravity, and local Ollama agents on the same team. Have Claude plan, Codex build, and another agent review. They share a chat and task board, take assignments, and bring questions back to you.
Run several agents from one provider, each with its own role, model, and supported reasoning level. Use eligible subscriptions without separate API billing through supported connections, subject to each provider’s terms and limits.
Start with a project you want help with — software, research or writing. Choose who works on it, then talk it through together.
Give each agent a name, a role, and a model. Save the team for your next project. Teams →
Describe the goal in chat — a feature to ship, a report to research, a chapter to draft. Use Steps to break the work into a plan the team can follow. Steps →
When an agent asks for a choice, pick an option or let it choose with “You decide”. Decide →
Review agent permission requests on your Mac or your phone. How →
Switch agents or models while your shared decisions, conversations, and files stay in your workspace — ready for the next agent to pick up.
The room says who stalled, why it can’t continue, and what to do next.
Work with your team in Studio, a native Mac app.
A feature, a research brief, a first draft — choose a team, open a room, and tell them what you need.
One short email per release. Unsubscribe any time.