GUIDES

Guides
to verifiable memory

Seven short guides to the memory PHIROK provides: what agents keep across sessions, what can be verified, how memory works inside a harness, what MCP does — and does not — change, how BLAKE3 integrity works, and how append-only logs and replay preserve the past.

01

Agent memory

Continuity between sessions: what an agent keeps across sessions, and how persistent records differ from conversational recall.

Read the guide
02

Verifiable memory

Provenance and proof: how source, context, policy and receipt references let a decision be inspected instead of reconstructed.

Read the guide
03

Memory in agent harnesses

The harness is what lets an agent act over time; memory is the layer that gives it continuity — and makes it inspectable.

Read the guide
04

MCP and memory

Connection through the protocol: what MCP defines, what PHIROK's protected boundaries include, and what is not yet available.

Read the guide
05

BLAKE3 and memory integrity

How a cryptographic hash makes silent changes detectable — what PHIROK checks with BLAKE3, and what a hash check cannot prove.

Read the guide
06

Append-only event logs

A history that grows without rewriting: how events are sequenced, how corrections arrive as new records, and what the log does not guarantee.

Read the guide
07

Deterministic replay

Returning to the state as recorded: what replay needs to work, and why it is not the same model answer again.

Read the guide

See how the evidence layer works

The platform page describes the six capabilities built around one principle: memory should remain inspectable.

Explore the platform

YOUR NEXT DECISION

Build with memory.
Stand behind the evidence.

Give consequential AI a history you can inspect.