Permem

Permem adds memory to LLM applications. It is built for developers who want an AI system to remember past conversations, preferences, and facts, and it describes this as automatic memory for any LLM.

The service automates what to store and when to store it, and it handles deduplication. Its workflow centers on two methods: inject() before an LLM call and extract() after the response. inject() retrieves relevant memories and formats them for a system prompt, while extract() analyzes the conversation, extracts facts, skips duplicates, and stores new memories. The product also says it can use memorize() and recall() for manual memory management or as tools for AI agents. Each memory includes type, importance, emotions, entities, topics, and timestamps, which are extracted automatically. Memories are linked into a knowledge graph, and connections can be visualized per user or across an entire project.

Permem is delivered as code for TypeScript and Python, with usage shown through an import from permem and an instantiated client. The page also presents Get Started, Login, Free, and View Docs links. Its own wording describes it as persistent memory for AI and as a retrieval and injection layer for context, but the page does not provide additional licensing or pricing details beyond the mention of Free.