Run Claude Code on twenty machines from one browser tab
A coding agent that edits files and runs commands needs someone to say yes. Run one in a terminal and you sit with it. Run twenty and you cannot.
One console instead of
What it does
The Console
One tab per live session, down the side of the window — across every machine you run.
Learn more →Approvals You Can Read
The agent asks in structured form, so the console can render the actual change rather than scraped text.
Learn more →The Destructive Tier
Two classifiers, one on the console and one on the machine. Where they disagree, the stricter reading wins.
Learn more →Machines Dial Out
Each machine opens a connection to the console. Nothing listens on the VM.
Learn more →Credentials and Installation
Register a machine, get a claim good once. The token comes into existence on the machine.
Learn more →Sessions That Survive
Networks drop and machines restart. A conversation should not end because of it.
Learn more →How it is built
Machines dial out
Each machine opens a connection to the console. Nothing listens on the VM, so there is no inbound port to open and no firewall rule per box.
Graded twice, independently
Every tool call is classified read-only, write or destructive by the console and again by the machine. Where they disagree, the stricter reading wins.
78 end-to-end tests
The console, the bridge and the machine installer are covered by browser tests that drive the real stack against a scripted agent.
Stop watching one agent at a time
Every session on every machine in one tab, every approval something you can actually read, and nothing waiting on you that you have to go looking for.
Get in touch