Memory that thinks.
Knowledge that lasts.
What is Docket?
Most knowledge systems treat memory as a flat vector database. Docket treats it as a living, structured memory graph: every fact has context, lifetime, importance, and relationships. You can ask complex questions across time, retrieve memories by salience, and build agents that remember what actually matters.
Why Docket?
RAG gives you answers from documents. Docket gives you a system that reasons like you do.
- ✓Stop losing context across sessions
- ✓Query what was true at any point in time
- ✓Control forgetting instead of brute-force context windows
- ✓Own your memory stack — self-hosted and open-source
Built for builders
Docket is designed as a pluggable core. Swap LLMs, embedding models, vector stores, blob providers, and queues without rewriting your application. Deploy it on your laptop, Cloudflare, AWS, or anywhere Node.js runs.
Core capabilities
Everything you need to build memory-backed applications.
Cognitive memory model
Temporal knowledge graph
Composite retrieval
Adaptive decay
RBAC access control
Pluggable adapters
Start building with Docket
Get up and running in minutes, then explore the full architecture when you are ready to extend it.