DCL Evaluator
Cryptographic audit trail for every AI agent decision
Start new thread
Dari Rinch

DCL Evaluator - Cryptographic audit trail for every AI agent decision

by•
Can you prove what your AI agent actually decided? DCL Evaluator gives you cryptographic proof of every LLM decision — deterministic, tamper-evident, bit-for-bit reproducible. Every output is evaluated against your policy. COMMIT or NO_COMMIT. Each decision gets a SHA-256 hash, chained to the previous one. Works with Ollama, Claude, GPT-4, Grok, Gemini. 100% offline. Desktop-first.

Add a comment

Replies

Best
Dari Rinch
Maker
📌
Hi Product Hunt! 👋 I'm Dari, indie developer from Siberia. I built DCL Evaluator after realizing there's a blind spot in every AI pipeline: you can log what an agent said, but you can't cryptographically prove it. Probabilistic guardrails help, but they're not reproducible — same input, different answer. DCL is different: deterministic policy engine + SHA-256 hash chain = tamper-evident audit trail that holds up under regulatory scrutiny. Happy to answer any questions about the architecture or use cases. And yes — it runs fully offline with Ollama. 🦾
Dari Rinch
Maker

Curious what's the biggest pain point for you when auditing agent behavior — is it reproducibility, policy compliance, or something else entirely?

Shitcoiner

@daririnch The real problem is the lack of provability. You can set up compliance, but how do you prove to a regulator or a client that the agent followed the rules in every single iteration?

Dari Rinch
Maker

@shitcoiner Exactly the distinction that matters: compliance is a promise, provability is a receipt.

DCL produces tamper-evident audit logs per decision cycle — TX-hash verified, independently anchored. A regulator doesn't need to trust the agent; they can inspect the trace.

That's the core design principle.