Show HN: Long-term memory for AI agents and teams, built with PostgreSQL
TL;DR
Hey folks! Over the past weeks, I started building a long-term memory for AI agents. While there are many around already, this one is specifically aimed at development teams. There are a few items which make it special: - Development teams can share memories between their agents (and across agent types such as Claude and Codex) - Higher value memories become artifacts, such as code style rules, decision, feature designs, ...
Nauti's Take
Shared memory across agent types is a real gap in the market — anyone running Claude, Codex and custom agents side by side knows the sync pain. Hierarchical scopes for code style and decisions sound workable, provided privacy and audit are solid.
Early-stage risk: a self-hosted PostgreSQL stack needs care; teams without ops muscle should pilot it behind a feature flag first.
Sources