Mohammed Arshad Ansari has open-sourced AEGIS, a self-hosted AI orchestration platform designed to streamline personal workflows.The system, built for a single user, automates repetitive tasks like email management, financial tracking, and knowledge organization.It employs four agents—Sebas (GTD), Raphael (research), Maou (money), and Pandora's Actor (infrastructure)—each with distinct responsibilities.

The platform uses FastAPI, Postgres (with pgvector), and Temporal for durable workflows, with LiteLLM proxying local-first models like Claude or GPT for complex tasks.

Key innovations include a unified interruption mechanism via a PostgreSQL table, eliminating per-feature approval tables, and database-driven agent capabilities instead of hardcoded logic.This design allows users to customize agents, rescope their functions, or add new ones without modifying code.

AEGIS emphasizes local-first computing, with tiered model usage (fast/balanced/smart) and explicit handling of reasoning token limits to prevent hidden costs.

While not a SaaS solution or a framework for others, it serves as a complete, opinionated application for personal automation, inviting users to fork, configure, and test its capabilities.