DCL Evaluator

Name: DCL Evaluator
Author: Fronesis-Labs

@Fronesis-Labs

Repository

About DCL Evaluator

DCL Evaluator — cryptographic audit trail for AI agents. Powered by Leibniz Layer™

#audit-cryptography-ai-agents-llm-compliance

Basic information

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "dcl-evaluator": {
      "command": "npx",
      "args": [
        "-y",
        "@smithery/cli@latest",
        "run",
        "dcl-evaluator--fronesislabs",
        "--key",
        "<YOUR_DCL_API_KEY>"
      ],
      "env": {
        "DCL_API_KEY": "<YOUR_DCL_API_KEY>"
      }
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is DCL Evaluator?

DCL Evaluator is the first implementation of Leibniz Layer™, a cryptographic verification protocol for AI agent decisions. It provides deterministic, tamper-evident audit trails for any LLM-powered system, sealing every decision into a hash chain for integrity verification.

How to use DCL Evaluator?

Connect via MCP by adding the server configuration to your client, providing the URL and an API key. Alternatively, download the Windows desktop app. Four tools are available: dcl_commit (seal an action), dcl_verify (verify chain integrity), dcl_get_chain (retrieve audit trail), and dcl_report (generate compliance report). Autonomous agents can pay for audits via the x402 HTTP protocol.

Key features of DCL Evaluator

SHA-256 hash chain seals every agent action cryptographically.
Merkle tree audit trail is tamper-evident by design.
Deterministic policy engine ensures 100% reproducible decisions.
Drift detection uses statistical Z‑test for behavioural changes.
Multi‑LLM support: Claude, GPT‑4, Grok, Gemini, DeepSeek, Ollama.
Compliance reports export as tamper‑evident PDFs.
Runs fully offline and air‑gapped with zero data leaving infrastructure.

Use cases of DCL Evaluator

Audit AI agent decisions for legal, financial, or regulatory compliance.
Verify chain integrity to detect post‑hoc tampering.
Generate tamper‑evident compliance reports for regulators.
Monitor agent behavioural drift over time.
Enable autonomous agents to request and pay for audits without human intervention.

FAQ from DCL Evaluator

What problem does DCL Evaluator solve compared to standard logging?

Most AI systems are black boxes with no tamper‑evident record. DCL Evaluator commits every decision into a hash chain; modifying any past record invalidates the entire chain, providing mathematical proof of integrity without trust.

What are the runtime dependencies?

The desktop app runs on Windows. It works fully offline and air‑gapped; no data leaves your infrastructure. MCP connection requires only an API key and internet access to the endpoint.

Where does audit data live?

Data stays entirely on your infrastructure when using the offline desktop app. When using the MCP endpoint, data is processed through the server; the README does not specify data residency for the cloud service.

What transports and authentication are used?

MCP uses SSE transport via URL https://mcp.fronesislabs.com/sse with an x-api-key header. Additionally, the x402 protocol uses HTTP 402 for payments; no API keys are needed for autonomous audit requests.

Are there any known limits?

The README does not list specific limits. It supports multiple LLMs and built-in policy templates including EU AI Act, GDPR, Finance, Medical, and Red Team.