Each bot is a named teammate with its own memory. They share one computer, run on a schedule, ask before doing anything risky — and hand work to each other.
npm install -g @anthropic-ai/claude-code
Each bot has a name, a job, a persona and a durable memory file it writes to itself. It remembers how you like things done.
Give one bot a coordinator role and it will hand work to the right teammate on its own, then check what comes back. You watch the whole exchange.
Every tool call is intercepted. Deleting, publishing, spending, pushing — all stop and wait for you. Allow once, always allow, or deny.
One folder your bots share, with a live activity feed, a file browser and a terminal you can drive yourself.
“Every weekday at 8:00am.” Routines run a bot on a schedule with run history, a pause switch and a test run.
Save a process as a reusable skill and call it with /name.
Any bot can use it.
Work keeps running with the window closed, but not with the PC off.
Bots get a terminal, files, web search and MCP connectors — not a mouse.
SmartScreen will warn on first run — it's an independent build with no code-signing certificate. “More info” → “Run anyway”.