Feature
Bi-Temporal Claims Ledger
Most memory systems store conversations as blobs and hope retrieval sorts out the truth. MetaMemory additionally distills memories into structured claims — atomic facts with provenance — and tracks each one bi-temporally: when the fact was true in the world, and when the system learned it. When a user changes jobs or a decision gets reversed, the new claim supersedes the old one instead of contradicting it at retrieval time.
Facts, Not Just Blobs
An extraction pipeline distills stored memories into claims: subject, fact, provenance, and timestamps. Retrieval can then inject "known current facts" directly into context, so agents answer from the current state of knowledge rather than whichever old conversation happened to rank highest.
Bi-Temporal Supersession
Every claim carries two time axes — valid time (when the fact held in reality) and system time (when MetaMemory learned or invalidated it). Updates supersede rather than overwrite: the ledger maintains exactly one live current fact per claim while preserving full history.
Point-in-Time Queries
Ask what was known as of any moment. The asOf API reconstructs the knowledge state at a given time, and the change feed reports which facts changed between two points — useful for audits, debugging agent behavior, and explaining why an agent said what it said.
Trust-Weighted Arbitration
In multi-agent fleets, claims carry agent attribution and per-agent trust levels. When sources conflict, a trust-weighted arbitration ladder decides which fact stands — and a contested claim never silently voids the established current one.
87.5% (LongMemEval)
Knowledge-Update Accuracy
Valid + system
Time Axes
Exactly one per claim
Live Current Facts
asOf + change feed
History Queries