Skip to content
Docs

Welcome

What Efficient Labs and the Atmosphere are, the sovereignty thesis, and the honest-status philosophy these docs are built on.

On this page

Efficient Labs builds sovereign AI infrastructure: your agents, your hardware, your rules. StratosAgent is the agent that runs on a machine you own; the Atmosphere is the peer-to-peer mesh that connects those agents to the world without a central server that can seize, censor, or surveil them.

The sovereignty thesis

Most AI runs on someone else's computer. That means someone else holds your keys, your context, and your kill switch. We invert that: compute and keys stay on hardware you control, peers connect directly over an encrypted mesh, and identity is content-addressed and post-quantum sealed. Nothing is exposed to the public internet — there are no inbound ports to attack.

  • Local-first — inference and secrets stay on your machine.
  • Peer-to-peer — nodes hole-punch to each other; no broker server in the middle.
  • Content-addressed — skills and receipts are identified by their hash, not a URL someone can swap.
  • Post-quantum — signatures and seals use ML-DSA-65 today, not someday.

The honest-status philosophy

These docs never claim a capability is further along than it is. Every feature inherits a status level from our public capability matrix, and that level is shown inline as a badge. The four levels are:

  • Live — running in production, exercised by tests.
  • Wired — built and connected into the daemon; hardening in progress.
  • Standalone — built and tested in isolation; live wiring is supervised.
  • Mock — scaffold or placeholder, explicitly NOT real yet.
See it yourself
The same matrix that drives these badges is published on the homepage. Check the live status board at /#status before you rely on anything.
Was this page helpful?
Last updated June 3, 2026