Episodic memory engine that captures, indexes, and replays agent experiences. Semantic vector search. Persistent across sessions. Your AI remembers everything.
Three operations that give your AI persistent, searchable memory across every conversation and session.
Store any interaction as an episode with role, content, metadata, and timestamp. Each episode is automatically embedded as a 384-dimensional vector for semantic search.
Query memory with natural language. Chronicle finds the most semantically relevant episodes using cosine similarity over vector embeddings. Returns ranked results with summaries.
Delete individual episodes or wipe an entire user's memory. Full GDPR-compatible data deletion. Your users control their data.
Every episode embedded with MiniLM-L6-v2 (384 dimensions). Semantic search finds relevant memories even when words don't match. Cosine similarity ranking.
SQLite-backed with WAL mode. Memories survive restarts, deployments, and migrations. Full state export and import for backup and portability.
Automatic summarization of recalled episodes. Contextual summaries ready for LLM injection. Multiple output formats for different agent frameworks.
Each user/agent gets isolated memory space. No cross-contamination. Track episode counts, manage capacity, and list all active users.
Built-in usage tracking per API key. Free tier with 1,000 operations/month. Stripe checkout integration for paid plans. Usage dashboard.
Bearer token authentication. Per-key and per-IP rate limiting. Configurable limits. Production-ready security out of the box.