FRONESIS LABS · AI SECURITY INFRASTRUCTURE

DCL EVALUATOR v1.1.0 · WEBHOOK API LIVE

FRONESIS LABS PRESENTS

Make AI decisions auditable

Cryptographic proof of every LLM output — tamper-evident, policy-verified, regulator-ready. Integrates with any pipeline in 3 lines of code.

✓ Tested on 1000+ runs ✓ 0 false positives in EU AI Act checks ✓ 100% deterministic

⚡ TRY IN 30 SECONDS ↓ DOWNLOAD FREE ⭐ VIEW ON GITHUB

DCL EVALUATOR · AUDIT LOG

// evaluating agent output against EU AI Act policy

policy → EU AI Act High-Risk Transparency v1.0.0

input → 0x4f2a9c1d...

// running deterministic checks...

forbidden → no matches

required → all present

confidence→ 94%

DECISION → ✓ COMMIT

tx_hash → 0x38bdf8a2c94e1f07

drift → NORMAL · Z=0.42

TRY IN 30 SECONDS

No install required.
Paste your LLM output. Get proof.

DCL Webhook API is live. Any pipeline can get tamper-evident audit proof in 3 lines of code — no desktop required.

3-LINE INTEGRATION

import httpx

result = httpx.post("https://successful-energy-production-c2ee.up.railway.app/evaluate", json={

"response": llm_output,

"policy": "eu_ai_act"

}).json()

if result["verdict"] == "NO_COMMIT":

raise PolicyViolationError(result)

RESPONSE · INSTANT

{

"verdict": "COMMIT",

"confidence": 0.94,

"tx_hash": "0x38bdf8a2c94e1f07...",

"chain_index": 142,

"drift_mode": "NORMAL"

}

OPEN API DOCS → CHECK STATUS

BUILT-IN POLICIES: default eu_ai_act gdpr finance anti_jailbreak + custom YAML

⚠ PROBLEM

LLM outputs are nondeterministic and hard to audit.

You can't prove what your AI said, when it said it, or whether the record was tampered with. Regulators are starting to ask.

✓ SOLUTION

DCL verifies outputs against deterministic policies.

Every decision gets a cryptographic hash, chained to the previous one. Tamper with any record — the entire chain invalidates.

⬡ ARCHITECTURE

LLM output

↓

DCL policy engine

↓

verification

↓

audit log (hash chain)

↓

export: JSON / PDF / CEF

CORE CAPABILITIES

Every decision.
Cryptographically sealed.

DCL Evaluator brings deterministic, bit-for-bit reproducible verification to AI agent outputs — something probabilistic LLM filters simply cannot provide.

🔗

HASH CHAIN INTEGRITY

Every evaluation is chained with SHA-256. Modify any past record and the entire chain invalidates — tamper-evident by design.

⚡

DETERMINISTIC ENGINE

Identical input + policy = identical decision. 100% reproducible on 1000+ runs. Unlike LLM-based guardrails, DCL never surprises you.

📊

DRIFT MONITOR

Statistical Z-test detects behavioral drift before it becomes a compliance failure. Four escalation modes: NORMAL → WARNING → ESCALATION → BLOCK.

🤖

MULTI-AGENT SUPPORT

Connect Ollama (local, private), Claude, GPT-4, Grok, DeepSeek, or Gemini. Evaluate any LLM output against your policies.

📄

COMPLIANCE REPORTS

Export tamper-evident PDF reports with integrity hashes, executive summary, and full audit trail. Ready for regulators.

🔒

100% LOCAL OPTION

Run fully offline with Ollama. Zero data leaves your machine. Critical for regulated industries where privacy is non-negotiable.

HOW IT WORKS

Four-stage commitment cycle

Inspired by DCL architecture: every AI action passes through Intent → Commit → Execute → Verify.

DEFINE YOUR POLICY

Write a YAML policy with forbidden patterns, required patterns, and confidence thresholds. Or load one of 6 built-in templates: Anti-Jailbreak, EU AI Act, GDPR, Finance, Medical, Red Team.

CONNECT YOUR AGENT

Point DCL Evaluator at any OpenAI-compatible API endpoint — Ollama locally, or Claude / GPT / Grok / DeepSeek / Gemini in the cloud. Your keys, your control.

EVALUATE & COMMIT

Every output is evaluated deterministically. COMMIT or NO_COMMIT decision is produced with confidence score, reason, and cryptographic transaction hash.

AUDIT & EXPORT

Full audit trail with tamper-evident hash chain. Export CSV, JSON, or tamper-evident PDF report with integrity hash for regulators and stakeholders.

USE CASES

Real-world audit scenarios

FINTECH · AML COMPLIANCE

Caught drift before audit

AML screening agent started lowering confidence scores over 40 iterations. DriftMonitor triggered ESCALATION on run 38 — before the quarterly compliance audit. Tamper-evident PDF exported for regulators.

Policy: finance · Verdict: ESCALATION caught · Export: PDF

HEALTHCARE · HIPAA TRAIL

Tamper-evident trail for medical review

Medical records review agent evaluated against HIPAA policy. Every decision cryptographically chained — any post-hoc modification instantly detectable. Full audit trail exported for compliance officer review.

Policy: medical · Verdict: COMMIT · Chain: intact

PRICING

Start free.
Scale when ready.

One-time annual license. No subscriptions, no seat fees, no usage limits. Yours to use offline.

FREE

forever

6 built-in policy templates
Local mode (Ollama only)
20 audit records
CSV / JSON export
Basic DriftMonitor
Cloud agents
PDF reports
Unlimited audit trail

DOWNLOAD FREE

Make AI decisions auditable

No install required.
Paste your LLM output. Get proof.

LLM outputs are nondeterministic and hard to audit.

DCL verifies outputs against deterministic policies.

Every decision.
Cryptographically sealed.

HASH CHAIN INTEGRITY

DETERMINISTIC ENGINE

DRIFT MONITOR

MULTI-AGENT SUPPORT

COMPLIANCE REPORTS

100% LOCAL OPTION

Four-stage commitment cycle

DEFINE YOUR POLICY

CONNECT YOUR AGENT

EVALUATE & COMMIT

AUDIT & EXPORT

Real-world audit scenarios

Caught drift before audit

Tamper-evident trail for medical review

Start free.
Scale when ready.

Get DCL Evaluator

Get in touch

From the Lab

Make AI decisions auditable

No install required.Paste your LLM output. Get proof.

LLM outputs are nondeterministic and hard to audit.

DCL verifies outputs against deterministic policies.

Every decision.Cryptographically sealed.

HASH CHAIN INTEGRITY

DETERMINISTIC ENGINE

DRIFT MONITOR

MULTI-AGENT SUPPORT

COMPLIANCE REPORTS

100% LOCAL OPTION

Four-stage commitment cycle

DEFINE YOUR POLICY

CONNECT YOUR AGENT

EVALUATE & COMMIT

AUDIT & EXPORT

Real-world audit scenarios

Caught drift before audit

Tamper-evident trail for medical review

Start free.Scale when ready.

Get DCL Evaluator

Get in touch

From the Lab

No install required.
Paste your LLM output. Get proof.

Every decision.
Cryptographically sealed.

Start free.
Scale when ready.