A local-first conversational memory agent that turns scattered work threads into organized, source-backed knowledge.
Important decisions, setup notes, debugging context, and project ideas get buried inside long Slack threads, mixed AI conversations, documents, and notes. Weeks later, nobody can find what was decided — so the same knowledge work gets done twice.
Instead of acting as a standalone chatbot, Asterism lives alongside the places where people already work and talk. Conversation-style messages flow into the system; the agent extracts focused memories, organizes them into topics and broader galaxies, and preserves each memory’s original source context.
When you ask a question, it answers only from stored memories — not from broad model knowledge or a single chat history. That grounding is the point: it builds a reusable memory layer across real project conversations, and if there isn’t enough stored context to answer, it says so plainly instead of inventing something.
Product teams, software engineers, founders, and researchers — anyone who needs to recover decisions, share context, and stop losing knowledge to the scroll. Built with a team at an AI Tinkerers SF hackathon.