VCAL Project
Enterprise add-on • Available for pilots

VCAL Audit

A self-hosted evidence retention and traceability service for AI Cost Firewall deployments. Reconstruct request lifecycles, retain structured operational evidence, and verify stored records through an authoritative SHA-256 hash chain.

Built for teams that need self-hosted AI traceability, retained operational evidence, and a foundation for future SIEM, SOAR, and compliance integrations.

Traceability

Reconstruct request lifecycles

Correlate request validation, guard decisions, cache processing, upstream calls, and terminal outcomes through a stable trace identifier.

Integrity

Verify retained evidence

Store accepted evidence in authoritative sequence order and verify a SHA-256 record chain to detect unexpected modifications.

Operations

Query, export, and observe

List retained evidence, reconstruct traces, export NDJSON, and monitor ingestion, persistence latency, authentication, and chain verification.

How it works

A durable evidence boundary for AI Firewall

AI Cost Firewall emits structured evidence during request processing and delivers it asynchronously to VCAL Audit. Audit validates each batch, assigns authoritative sequence numbers, persists new events atomically, and updates a tamper-evident record chain.

Application

Chatbot, RAG, agent, or API client

AI Cost Firewall

Emits correlated evidence events

VCAL Audit

Validates, sequences, and persists

SQLite

Durable evidence and record chain

Example trace: a guarded chat request is received, scanned, served from cache or sent upstream, and completed.
Correlation
trace_id: stable across stages
Lifecycle
request.received → request.completed
Integrity
sequence + previous hash + record hash
Audit capabilities

Retain, investigate, verify, and export

VCAL Audit provides a compact evidence service for pilots and production-oriented single-node deployments.

Ingest

Accept authenticated evidence batches with schema validation, atomic persistence, and duplicate-event handling.

Investigate

List events in sequence order and reconstruct complete request lifecycles by stable trace_id.

Verify

Recalculate the authoritative SHA-256 record chain and identify the first invalid sequence if verification fails.

Export

Export retained traces as NDJSON for investigation, archival, and future SIEM or SOAR integration.

Deployment model

Self-hosted, compact, and independently deployable

VCAL Audit runs as a separate service close to AI Cost Firewall. The current implementation uses SQLite with write-ahead logging and full synchronous commits, making it suitable for pilots and controlled single-node deployments that prioritize durable, ordered evidence.

Pilot deployment includes

  • • Self-hosted Audit service
  • • AI Cost Firewall buffered delivery integration
  • • API-key protected evidence endpoints
  • • SQLite evidence persistence
  • • Health, readiness, version, and metrics endpoints
  • • Grafana dashboard for Audit operations
Use cases

Where VCAL Audit helps

Incident reconstruction

Rebuild the path of a request across validation, guards, cache activity, upstream calls, and terminal outcomes.

Guard decision review

Correlate Security Guard and Privacy Guard decisions with the surrounding request lifecycle.

Operational evidence retention

Retain metadata-first evidence independently of application logs and short-lived observability data.

Integrity verification

Verify whether stored records and their authoritative sequence remain consistent with the Audit hash chain.

Evidence export

Export traces as NDJSON for controlled investigation, archival, and downstream processing.

Compliance preparation

Create a traceability foundation for future compliance controls, reporting, and SIEM or SOAR integrations.

Enterprise governance direction

The evidence layer for broader AI governance

VCAL Audit complements AI Cost Firewall, VCAL Privacy Guard, and VCAL Security Guard by retaining structured evidence outside the request path. It establishes the traceability layer needed for investigation, future compliance controls, and integration with enterprise security operations.

FAQ

Is VCAL Audit available now?

It is available for enterprise pilots. General public self-serve distribution is not the primary model at this stage.

Does it send evidence to VCAL?

No. The intended deployment model is self-hosted inside your own environment, close to AI Cost Firewall and your application traffic.

Does it store prompts and responses?

AI Firewall's current evidence events are metadata-first and do not normally contain prompt or response bodies. Producers should still control free-form attributes to prevent unnecessary sensitive data retention.

Is the evidence immutable?

The stored records are tamper-evident through an authoritative SHA-256 hash chain. The current implementation does not yet provide external immutable anchoring.

Contact VCAL

Interested in a VCAL Audit pilot?

Contact VCAL to discuss evidence requirements, retention scope, AI Cost Firewall integration, deployment architecture, security review, pilot metrics, or enterprise packaging.