Agent Memory Graph

Documentation status: guide — see Maturity and evidence.

The agent memory graph stores what an agent observes, decides, attempts, validates, rejects, and learns.

Core concepts

define .:#Agent(
    identity:#Identity,
    role:#Role,
    state:#AgentState
);

define .:#MemoryItem(
    owner:#Agent,
    content:#Fact,
    context:#Context,
    source:#Source
)[
    confidence::#float
];

define .:#AgentRun(
    agent:#Agent,
    task:#Task,
    context:#Context
)[
    status::#RunStatus
];

A logiCells agent does not only remember text. It records structured relations that can be queried, constrained, reasoned over, and audited.