# Using djzs-trust with Claude

*Query ProofOfLogic certificates and agent trust scores directly from your AI assistant.*

By [Username in a Box](https://username.box) · 2026-04-03

---

    
    // SYS_ID: djzs-mainnet-01
    // LOGIC_TAXONOMY: PAYLOAD_AUDIT::MCP_DOCUMENTATION
    // STATUS: INGESTION_COMPLETE

What is djzs-trust?
-------------------

djzs-trust is an MCP (Model Context Protocol) tool that connects Claude directly to the DJZS Protocol — a blockchain-based adversarial logic auditor for autonomous agents. Once connected, Claude can query permanent audit records, verify reasoning traces, and check agent trust scores on Base Mainnet without leaving the conversation.

Every time the DJZS oracle audits a strategy memo, it writes a ProofOfLogic (PoL) certificate to Irys Datachain — permanently and publicly. djzs-trust gives Claude a live query interface into that certificate history.

* * *

Setup — 4 Steps
---------------

**Step 1: Open Claude Desktop settings**

Go to Settings → Developer → Edit Config. This opens `claude_desktop_config.json`.

**Step 2: Add the MCP server**

Paste this into the config file:

    {
      "mcpServers": {
        "djzs-trust": {
          "command": "/Users/YOUR_USER/.npm-global/bin/mcp-remote",
          "args": ["https://djzs-trust-mcp.easy-less-spoil.workers.dev/mcp"]
        }
      }
    }
    

Replace `YOUR_USER` with your Mac username.

**Step 3: Restart Claude Desktop**

Quit completely (Cmd+Q) and reopen. Go to Settings → Developer to confirm djzs-trust shows **running**.

**Step 4: Enable in conversation**

Click the + button in any Claude chat → Connectors → toggle djzs-trust on. You're ready.

* * *

5 Use Cases
-----------

**01 — Pre-execution audit check**

Before executing a high-stakes action — deploying a contract, releasing funds, or delegating a task — ask Claude to verify the audit history first.

> _Before we deploy this contract, query djzs-trust for any FAIL verdicts on ProjectName_

**02 — Agent due diligence**

Check the full audit record of any agent address before trusting it with work or capital. Every verdict is permanent and tamper-proof.

> _Query djzs-trust for all ProofOfLogic certificates for 0xAGENT\_ADDRESS_

**03 — Treasury-tier risk screening**

Filter specifically for high-value audits that failed. Treasury-tier audits are exhaustive stress-tests — a FAIL here is a hard stop.

> _Show me all treasury-tier FAIL verdicts from DJZS_

**04 — Certificate verification**

Verify any specific ProofOfLogic certificate by its Irys transaction ID. Confirms it's real, unaltered, and permanently stored.

> _Verify this Irys certificate: 71oNMzL4hgLoXo7SNEsgPSJ8oCETs15jKwioke3V2rSH_

**05 — Pre-payment escrow check**

Before approving a payment or releasing escrow to an agent, run a trust check. This is Audit-Before-Act in practice.

> _Before I approve this payment, check if this agent has any DJZS FAIL flags: 0xAGENT_

* * *

What Claude Returns
-------------------

When djzs-trust finds certificates, Claude returns a structured summary like this:

    {
      "total": 3,
      "pass": 2,
      "fail": 1,
      "certificates": [{
        "irys_id": "71oNMzL4hg...",
        "irys_url": "https://gateway.irys.xyz/71oNMzL4hg...",
        "verdict": "PASS",
        "tier": "micro",
        "target_system": "MyDAOProject",
        "timestamp": 1743000000
      }]
    }
    

The `irys_url` is a permanent public link — anyone can verify the certificate without authentication, forever.

* * *

Failure Taxonomy
----------------

When an audit returns FAIL, the certificate includes one or more DJZS-LF failure codes. Agents must halt on Critical flags.

Code

Name

Severity

Auto-abort

DJZS-S01

Circular Logic

Critical

Yes

DJZS-S02

Missing Falsifiability

Critical

Yes

DJZS-E01

Confirmation Tunnel

High

Yes

DJZS-E02

Authority Substitution

High

Yes

DJZS-I01

Misaligned Incentive

Medium

Review

DJZS-I02

Narrative Dependency

Medium

Review

DJZS-X01

Unhedged Execution

Critical

Yes

* * *

Available Tools
---------------

**query\_pol\_certificates**

Queries Irys Datachain for ProofOfLogic certificates. Filter by project name, wallet address, verdict (PASS/FAIL), or audit tier. Returns up to 100 results.

Parameters: `targetSystem`, `verdict`, `tier`, `limit`

**query\_agent\_trust**

Queries on-chain trust scores, fail rates, staking positions, and LF code history for DJZS-registered agents on Base Mainnet. Returns a PROCEED or HALT safety verdict automatically.

Parameters: `agentAddress`

* * *

Run Your First Audit
--------------------

The MCP queries certificates that already exist on Irys. To generate your first ProofOfLogic certificate, run an audit through the DJZS oracle at [**djzs.ai**](http://djzs.ai).

Once the audit completes, come back to Claude and query djzs-trust — your certificate will be there, permanently.

* * *

_DJZS Protocol —_ [_djzs.ai_](http://djzs.ai) _No agent acts without audit._

---

*Originally published on [Username in a Box](https://username.box/using-djzs-trust-with-claude)*