Providers lend the spare capacity of their machines; customers submit container tasks; a coordinator schedules them, meters the real usage and bills it per second. Machines and customers only connect out, so nothing is opened inbound.
Click the diagram to zoom. Source: docs/diagrams.
![]() |
![]() |
![]() |
![]() |
Top: customer overview and task view. Bottom: provider overview and add a machine.
A lazy cake is the no-bake chocolate biscuit cake made from leftover pieces you already have: no oven, no effort, and better than its parts. LazyCake does the same with computers: lots of small, idle ("lazy") machines combined into something useful with almost no effort.
- Run any container on spare machines, with real CPU, memory and disk limits.
- Pay per second for what is actually used; providers are credited for it.
- Private services reachable through gateways over an encrypted tunnel.
- Fair scheduling: trust scores, canary tasks and a three-way byte check against cheating.
- Dashboard, CLI and AI apps (through MCP) for customers and providers.
- TLS everywhere, with a self-renewing Let's Encrypt certificate.
Quickstart, public images, screenshots and design decisions · Architecture index · Changelog





