Memori

Why we included this project

Agents that need to remember what happened in earlier sessions usually end up with a hand-rolled memory hack. Memori is a ready-made layer for that: it watches each conversation turn, sorts what it sees into facts, preferences, rules, and summaries, and pulls back only the pieces a later prompt actually needs. It is agnostic to the LLM, the datastore, and the agent framework, so it slots into an existing stack without a redesign. Two things stand out for production use: recall is explainable, meaning every returned memory carries a reason it was included, and it saves tokens by retrieving concise structured state instead of dumping full history into the prompt. You can start on the managed cloud with an API key and move to your own VPC or on-premises setup later.