AI moves forward.
Evidence stays.
Verifiable memory for AI in production.
Preserve what was known. Inspect what was recorded.
demo_rct_7v2a…e91c Bound to the recordThe source stays connected to the decision.
Inspect the sample01 / THE QUESTION BEHIND THE ANSWER
An answer is a moment.
Evidence gives it a history.
When an AI makes a consequential decision, the answer alone is not enough. What source did it use? Which policy was active? What context was available?
We built PHIROK to keep those references together, so the past can be inspected instead of reconstructed from memory.
Understand the architecture“Escalate this case.”
The outcome, without its context.“Escalate this case.”
- Source
- Customer issue record
- Context
- Account tier & case history
- Policy
- Escalation v3.2
- Receipt
- demo_rct_7v2a…e91c
02 / THE EVIDENCE LAYER
One record. A complete chain of context.
Six capabilities built around one principle: memory should remain inspectable.
Evidence-first history
Record the source, context and configuration behind consequential AI outcomes.
Explore capability RCT-02Cryptographic receiptsConnect stored evidence through BLAKE3 hashes and traceable receipt references.WAL-03Append-only recordsAdd new evidence without silently rewriting what was recorded before.TNT-04Tenant separationKeep evidence scoped to the authenticated workspace it belongs to.GRH-05Traceable retrievalRetrieve evidence and its relationships with non-generative GraphRAG.RPL-06Replay & reviewReconstruct stored memory state and prepare structured evidence exports.No matching capabilities. Try “receipt”, “tenant” or “replay”.
6 capabilities
03 / BACK TO THE MOMENT
Return to the state.
Not a retelling of it.
Move through this sample history. Each checkpoint shows the context recorded at that moment.
Evidence enters the record.
A customer issue is captured as source evidence. The later decision has not yet been recorded.
- Source
- Customer issue record
- Context
- Not yet bound
- Decision
- Not yet recorded
Replay concerns stored memory state, not a guarantee of identical model output.
THE AGENT ECOSYSTEM
Memory for the agent era.
The agents you already use, named as they are. PHIROK is built to be their memory layer: what it records is signed, scoped to its tenant, and reproducible on demand. No black box: what your agents remember becomes evidence you can audit.
YOUR NEXT DECISION
Build with memory.
Stand behind the evidence.
Give consequential AI a history you can inspect.