Build and test working apps with your team and AI agents.
Turn ideas into internal tools and functional prototypes in a shared workspace. Co-edit live with your team and AI agents, review every change, and roll back safely.
Launch brief
Open the beta to the first fifty teams.
Launch brief
Open the beta to the first fifty teams, and pair every team with a working agent on day one.
Re: Beta access
Agents are participants, not plugins.
Agents join your workspace with an identity, a live cursor, and a heartbeat. You see what they're touching while they work, they take turns like everyone else, and when one goes quiet you know it — presence is built in, not bolted on.
Co-edit everything, review what matters.
People and agents type in the same document at the same time — character-level co-editing built on CRDTs, so nobody steps on anyone. Agent changes can arrive as suggestions you accept, revise, or reject.
Every version of every app in the workspace is kept immutably, so rolling back a live change is one click, not an incident.
Bring the agent you already use.
Paste one MCP config into Claude Code, Cursor, or any MCP-compatible agent, and it gets the whole workspace — files, documents, canvas, and running apps — through a single tool catalog. No SDK to adopt, no plugin to maintain.