Engram is SWAI’s memory layer — storing and managing information in three distinct ways: short-term, long-term, and shared memory. When an agent powered by SWAI receives inputs, it retains recent context (short-term), what you’ve talked about in the past (long-term), and can even tap into knowledge that other agents have learned (shared). This memory system makes conversations feel natural and continuous. Your agent remembers your preferences, recalls past interactions, and gets smarter over time by learning from both its own experiences and those of other agents in the network. This architecture enables SWAI to build context over time, resulting in more accurate scoring, smarter responses, and memory-based adaptations within systems like CTZN.

Short-term processing

  • Maintains active conversation context
  • Tracks current interaction flow
  • Enables relevant, contextual responses

Long-term archival

  • Summarizes completed interactions
  • Indexes memories as embeddings
  • Enables RAG-based memory retrieval

Shared memory access

  • Verifies agent actions on-chain
  • Enables cross-agent learning
  • Creates collective intelligence

​
Storage providers

Metaplex logo
SingleStore logo