DCL Trust Oracle seals every AI agent decision into a tamper-evident chain — so when something goes wrong, you don't have to take the agent's word for it. Verify locally, in your browser, right now.
A real hash chain running in your browser. Simulate an agent decision, or corrupt a hash and see verification fail.
| Index | Tx Hash | Prev Hash | Verdict |
|---|
Hashes are real, computed client-side over each block's contents. Nothing here is sent anywhere — it's the same chaining logic the live server uses, running locally so you can see it break.
import { DclClient } from '@fronesis-labs/dcl-sdk'; const dcl = new DclClient(); const isValid = await dcl.verifyChain(); // Returns false if corrupted!
DCL is a protocol first — this site is its production implementation.
hash + metadata, never raw content. Any implementation can follow the same rules this server does.npm install @fronesis-labs/dcl-sdk.No subscription, no tiers — you pay for exactly what you call, in USDC via x402.
Point any MCP-capable client at the live server. x402-aware clients pay automatically per call.
{
"mcpServers": {
"dcl-trust-oracle": {
"url": "https://mcp.fronesislabs.com/mcp"
}
}
}
Payment is handled via x402 (USDC on Base). No API key, no signup — just a wallet that can settle per call.
Nothing raw is stored. Every audit-chain entry holds a hash and metadata only — never the text that was evaluated.
Prefer a free, offline first pass? Use the instruction-only checklists in DCL Skills before calling the live server.