Introducing SurrealDB Agent Memory

SurrealDB has introduced Agent Memory, a persistent memory and knowledge layer for AI agents that converts conversations, documents, and structured inputs into queryable entities, relationships, facts, embeddings, provenance, and execution traces stored within SurrealDB. Designed to address limitations of context windows and vector-only retrieval, it reconciles new information with existing records so agents can distinguish current facts, historical facts, uncertain claims, and explicitly removed information, supported by a tri-temporal model that tracks real-world validity, system knowledge, and database transaction history. The product combines document and conversational knowledge in a single graph-backed database and supports hybrid retrieval using vector search, full-text search, graph relationships, and structured lookups, with hierarchical scopes and isolated Contexts for multi-user or multi-tenant applications. SurrealDB reports a 77.2% mean score and 57.1% strict accuracy on the BEAM 1M long-term memory benchmark, while documenting methodological differences from the reference setup and identifying event ordering as a weaker area. Agent Memory is available through REST, MCP, Python and TypeScript SDKs, and integrations with frameworks including LangChain, OpenAI Agents, and Vercel AI SDK, with management and inspection tools provided through SurrealDB Studio.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.