Platform
SUBCORTEX
Cognitive Infrastructure for AI Agents
A persistent, portable cognitive layer — memory, identity, and belief that outlive any single model or vendor. Agents that remember what happened, know who they are, and can be held to what they said.
Capabilities
The layer underneath
Models are becoming commodities. What persists — memory, identity, accountability — is where the durable value sits.
Persistent Memory
Agents remember what happened across sessions, models, and vendors. Memory that outlives the conversation it was formed in.
Portable Identity
An agent's identity is unbundled from the model running it. Take the same agent anywhere MCP reaches.
Bi-Temporal Fact Engine
What was believed, and when it was believed, are tracked separately. You can reconstruct the state of knowledge at any past moment.
Accountable Belief
Every assertion carries provenance. When an agent acts on a belief, you can trace exactly where that belief came from.
Tenant Isolation
Fail-closed authorisation at the chokepoint. Cognitive data is segregated by tenant as a structural property of the engine itself.
Agent-to-Agent Context
A shared cognitive substrate means agents can hand off work with the context intact, rather than re-deriving it from scratch.
Give your agents a memory
SubCortex is available to design partners building agentic systems that need to remember. Tell us what you're building.