Write Sealed
An agent decision is committed to a verifiable chain of custody.
CORTEX Persist is the sovereign trust substrate that wraps agentic memory with cryptographic lineage, integrity proofs, and zero-loss auditability.
An agent decision is committed to a verifiable chain of custody.
The record verifies cleanly until an out-of-band mutation appears.
A review-ready audit pack is generated for humans who need proof.
$ pip install cortex-persist
$ cortex init --ledger main
$ cortex store "agent:refund_approved_150usd"
$ cortex verify --integrity-check
$ sqlite3 cortex.db "update ledger set value='tampered'"
$ cortex audit --taint-detection
> STATUS: CRITICAL_TAMPER_DETECTED
> EXPORT: artifacts/audit_v1.proof Built for:
CORTEX Persist sits on top of the memory layer you already use and turns agent activity into defensible evidence.
Sit above your current memory layer, vector database, or runtime without forcing a full rewrite.
Create cryptographic lineage so records can be verified later instead of trusted on faith.
Surface manual edits, missing records, and broken chains before audits or incidents uncover them.
Generate review-ready evidence for incidents, regulated workflows, and enterprise sign-off.
Use the Python SDK and CLI to get from first write to exported evidence without rebuilding your stack.
from cortex_persist import CortexClient
client = CortexClient(api_key="ctx_...")
client.memory.store("order-101", "refund_auth:true")
# Proof-of-Lineage Verification
proof = client.trace.get_ledger_proof("order-101")
print(f"Verified: {proof.is_intact}") One happy path, one tamper event, and one artifact your team can actually review.
Create a local ledger for one agent workflow.
Write a decision, event, or state mutation.
Confirm the chain of custody is intact.
Change the underlying record out of band.
Flag the mismatch immediately.
Produce an audit pack for review.
Lead with compliance-grade audit trails, then expand into incident forensics and enterprise rollout.
Prove who decided what, when, and from which state in regulated agent workflows.
Request WalkthroughInvestigate agent failures with verifiable evidence instead of fragile log reconstruction.
Request WalkthroughGive security, risk, and legal teams a concrete artifact before production rollout.
Request Walkthrough