From script execution and visual workflows to a public docs assistant powered by retrieval-augmented generation
Published: July 4, 2026
I built Autosage, a remote automation platform that lets users write scripts, orchestrate them as visual workflows, execute them on Linux and Windows machines, trigger them manually or automatically, and monitor everything through real-time logs.
On top of that, I built Autobot, an AI assistant that can generate scripts and workflows, help investigate failed runs, and power a public documentation assistant embedded in the docs site.
At a high level, Autosage is split into three planes:
And alongside those, a dedicated FastAPI AI service called Autobot that handles chat, tool use, execution copilot flows, and the public docs assistant.
Autosage came from a very practical problem: remote automation is usually fragmented.
You may have scripts living in random folders, cron jobs hidden on individual servers, webhook glue code scattered across services, and almost no clean way to visualize what runs where. Even when the automation works, the operational experience is messy.
I wanted one place where I could: