Agent memory you can
erase — and prove it’s gone.
Verifiable forgetting for AI-agent memory. It remembers what you tell it about a subject — and forgets on command, then proves the erasure: it existed, it's gone, it's irreversible.
The real thing builds this from your records — open the app for the full graph.
A real query against the live graph — not a script. For the full Forget & Prove flow, open the app.
Watch it forget.
Memory you can rely on.
Persistent threads
Every conversation saved — pick it back up later.
Isolated workspaces
Separate knowledge bases, each with its own graph.
Multi-turn context
Follow-ups understand what you just asked.
Runs fully offline
Point it at a local model — nothing leaves your machine.
Verifiable deletion
One atomic erasure across documents, graph and key — with a 3-part proof and a signed certificate.
Hybrid memory
A knowledge graph and vector embeddings in one CockroachDB store.
Proof of prior existence
AS OF SYSTEM TIME reconstructs exactly what existed the instant before erasure — from the database's own history.
Crypto-shredded keys
Each subject's data key is destroyed on forget, so residual ciphertext — MVCC history, backups, S3 — is unrecoverable.
Memory health, measured
A deterministic staleness score plus reversible demote / restore — free, 0 tokens.
The whole loop — including the verb everyone skips.
Every step is a real call in the codebase — remember, recall, and the one almost no one ships: forget.
- ingestRememberstore encrypted
- extractStructuregraph + embeddings
- askRecallgraph-grounded answers
- forgetForget & Prove the differentiatorverifiable erasure
- certificateCertifysigned & WORM-stored
Forgetting, proven.
Most memory products claim they handle stale knowledge. We measure it — decommission three systems, re-ask, and let a blind judge score every answer.
Everyone builds AI that remembers more.
The whole field accumulates — and stale memory misleads as often as it helps. Obliviate makes the opposite move: forgetting a subject, on command and verifiably.