How Houston runs in the cloud.
The Houston desktop app puts one agent on your machine. Houston Cloud puts ten thousand agents on a cluster, each one in its own sealed box, woken on demand, billed only when it works. This is how every piece of that stack fits together, written so a non engineer can follow it.
Reading legend
Each chapter has a status pill. Four states.
- Shipped already in the codebase today.
- Partial some of it works, the rest is on the plan.
- Planned proposed design, not built yet.
- Gate plan sequencing, decisions, and what triggers the build.
The big idea, in one paragraph
Houston Cloud is the same houston-engine from the
desktop, packaged in a tiny Linux box, and run one box per agent
on Kubernetes. Each box is sealed off from the others using
Firecracker microVMs. When nobody is talking to an agent, the
box sleeps and costs nothing. When a message arrives, the box
wakes up in under a second, does the work, and goes back to
sleep. A small control plane in front of it all handles login,
permissions, and routing messages to the right box. That's
it. The whole guide is just zooming into each of those words.