I'm Florian, a keen observer in the Open/Nano ...claw space. I programmed AutosClaw as an exercise in understanding the trust models that these personal assistants use.

This AutosClaw features:

* All agents running in ephemeral Docker container (like nanoClaw)

* Agents can have the ability to spawn asynchronous agents

* Project based secrets (Github tokens), that can be inherited by all agents in that project (and their spawned agents)

* A real-time dashboard with all running/deleted agents with full history

* A fast reload directly from UI (enables an agent PR -> Github repo review -> direct reload workflow)

* A chat view, in which you can directly interact with any agent orchestrated on the platform.

It is implemented in 8,017 lines of Typescript code (backend + frontend)

This is my first Show HN, I hope someone out there finds it useful :)