📋 Full Audit Trail

Every memory operation logged. Every decision traceable. Every data point has provenance.

In regulated industries, "trust me" isn't enough. UAML maintains a complete, tamper-evident audit trail for every operation on your AI agent's memory — who wrote what, when, why, and what reasoning led to each decision.

What Gets Logged

✍️ Write Operations

Every memory creation, update, and deletion is logged with timestamp, actor identity, source context, and the data before and after the change. You can reconstruct the complete history of any memory entry.

🔍 Read Operations

Memory access is logged too — who queried what, when, and what was returned. This supports data access audits and helps detect unauthorized access patterns.

🧠 Reasoning Traces

When your AI agent makes a decision based on recalled memories, UAML captures the reasoning trace: which memories were consulted, how they were ranked, and what logic chain produced the final answer. Full explainability.

Data Provenance

🔗 Source Tracking

Every memory entry records its origin: where did this knowledge come from? A meeting transcript, a document, a user message, an API response? Provenance metadata travels with the data, enabling trust verification at any point.

🔏 Tamper Evidence

Audit logs are append-only and cryptographically chained. Each entry includes a hash of the previous entry, creating a verifiable chain. Any tampering with historical records is detectable.

Compliance Queries

from uaml.audit import AuditLog log = AuditLog() # Who accessed customer data in the last 30 days? accesses = log.query( data_class="customer", operation="read", since="30d" ) # Show the reasoning trace for a specific decision trace = log.reasoning_trace(decision_id="dec-2026-0142") print(trace.chain) # Full logic chain with source references # Export audit log for external review log.export("2026-Q1-audit.json", format="json", signed=True) # Verify log integrity assert log.verify_chain() # True if no tampering detected

Why It Matters

← Back to UAML