Memory System Overview
How AgentOS memory works end-to-end — classifier-driven dispatch, BM25 + dense + Cohere rerank, Ebbinghaus decay, consolidation — and how each piece earned its place on the 85.6% LongMemEval-S / 70.2% LongMemEval-M result.
Memory Architecture Overview
How the three memory layers compose — standalone Memory, AgentCognitiveMemory, and the CLI — plus cognitive science foundations and competitor comparison.
Cognitive Memory
Cognitive memory for AI agents: Ebbinghaus decay, HEXACO-modulated encoding, Baddeley working memory, ACT-R spreading activation, 8 neuroscience-grounded mechanisms. 85.6% on LongMemEval-S, 70.2% on LongMemEval-M.
Cognitive Mechanisms
How AgentOS implements reconsolidation, retrieval-induced forgetting, temporal gist, schema encoding, source-confidence decay, and emotion regulation
Working Memory
A human-readable .md file that persists across conversations — inspired by Mastra's agent notepad pattern. Complements the Baddeley cognitive working memory with durable, editable state.
Memory Auto-Ingest
After every conversation turn, the auto-ingest pipeline uses a cheap LLM (gpt-4o-mini or claude-haiku) to extract structured facts from the exchange and store them in the agent's vector database. This runs non-blocking so it never slows down the conversation.
Agent Memory Tools
Reference for the 6 agent-facing memory tools (memory_add, memory_update, memory_delete, memory_merge, memory_search, memory_reflect) that let agents manage their own long-term memory.
Memory Import/Export
Export agent memories to SQLite, JSON, Markdown, or Obsidian vaults, and import from SQLite, JSON, Markdown, Obsidian, ChatGPT exports, and CSV.
Memory Scaling (4-Tier Vector Storage)
How to scale AgentOS retrieval from a single-agent SQLite file to billions of vectors across Postgres and Qdrant.
SQL Storage Quickstart
One storage adapter for AgentOS that runs on better-sqlite3, sql.js, Postgres, Supabase, Capacitor SQLite, and IndexedDB — automatic backend detection, schema parity, cloud backups