Brain: persistent knowledge
When you connect your sources, Brain ingests them: documents, structured data and internal references, chunked and indexed for governed retrieval, encrypted at rest and isolated to your organization.
Knowledge & Memory
Your knowledge stays with your models: Brain, Memory and Knowledge Intelligence arrive on roadmap V2: Brain persists organizational knowledge, Memory keeps the context your users, agents and tasks need, and Knowledge Intelligence proposes what becomes authoritative knowledge only after human review.
Brain and Memory are separate capabilities with separate budgets, both coming soon (roadmap V2): Brain will hold what the organization knows; Memory will hold what a session, user or agent is working on.
When you connect your sources, Brain ingests them: documents, structured data and internal references, chunked and indexed for governed retrieval, encrypted at rest and isolated to your organization.
Memory is context attached to users, agents or tasks; it can be persistent per user or agent, or session-scoped. Retention and access follow your organization policies.
Your systems will plug in through MCP servers and your own tools; retrieval, access control and citations will apply to connector data with the same guardrails as native capabilities.
Discovery is a candidate pipeline, never an automatic publication.
Incoming knowledge is classified, summarized and deduplicated against what Brain already holds.
Candidates are related to existing knowledge and screened for contradictions before anyone reviews them.
Candidates stay candidates: human review and organizational policy decide what becomes authoritative Brain knowledge.
Bring the sources your AI must know and the context your agents should remember, both governed like the rest of the control plane.
Region-pinned inference across the live model catalog, under one SLA.
Build agents with tools, MCP access and multi-step execution.
Authenticate the right people and machines for every request.