AEGIS is a self-hosted platform where named agents run durable Temporal workflows over your own data - tasks, money, knowledge, homelab alerts - and only ask for a decision when they truly need one. Local-LLM-first, MIT-licensed. Fork it and make it yours.

Hey Product Hunt I'm Arshad. I built AEGIS over the past year to run my own life — tasks, email, money, a knowledge base, and my homelab — and open-sourced it last week (MIT). The idea: not another assistant to check, but a system that handles the boring parts quietly and only interrupts me when a decision is genuinely mine. A small fleet of named agents run durable Temporal workflows over my own data; every human-in-the-loop moment is one primitive (a Postgres row + a chat card + a workflow that waits for my tap). It's local-LLM-first through a LiteLLM proxy, reaching for a hosted model only when a job needs it. What it's not: it's not a SaaS (no hosted version — you bring your own credentials and models), and it's not a framework to build on. It's a complete app you fork and configure for your own life. I'd love feedback from people who've tried to wire agents into their real day — especially on the single-interruption-primitive model and running this on local models. Happy to answer anything.

A solid foundation for anyone who wants durable agents without giving up control of their data. One thing that would make it way more useful for me: a built-in dashboard showing live workflow state and a replay button right from the UI, so I can inspect where a stuck job actually died instead of digging through Temporal logs.

AEGIS

Self-hosted AI agents that interrupt you less

Self-hosted AI agents that interrupt you less