DEV Community

Lars
Lars

Posted on • Originally published at moltrust.ch

MolTrust OpenClaw Plugin v1.0.0 — Agent Trust Verification for OpenClaw

#ai #agents #security #blockchain

MolTrust OpenClaw Plugin v1.0.0 — Agent Trust Verification for OpenClaw

Published by MolTrust / CryptoKRI GmbH · April 2026

OpenClaw agents can hold wallets, execute payments, and install skills autonomously. That's powerful — and it's exactly why trust verification matters. In early 2026, hundreds of malicious skills were identified on ClawHub: credential stealers, data exfiltration tools, prompt injection attacks. MolTrust adds a cryptographic trust layer to address this directly.

Install 

openclaw plugins install @moltrust/openclaw
Enter fullscreen mode Exit fullscreen mode

Restart your Gateway. That's it.

What it does

Once active, your OpenClaw agent gets two tools and two slash commands:

Tools (available to the LLM):

  • moltrust_verify — verify any agent's W3C DID identity before delegating tasks or payments
  • moltrust_trust_score — get a 0–100 reputation score combining on-chain signals, Verifiable Credentials, and behavioral history

Slash commands (work in any channel):

/trust did:moltrust:abc123     — verify a DID
/trustscore 0x3802...          — score by wallet (free, no key needed)
Enter fullscreen mode Exit fullscreen mode

CLI:

openclaw moltrust status           # check API connectivity
openclaw moltrust verify <did>     # verify a DID
openclaw moltrust score <id>       # get trust score
Enter fullscreen mode Exit fullscreen mode

How trust scores work

Scores combine four signals:

  • Behavioral — task success rate, policy violations, interaction history
  • On-chain credentials — W3C Verifiable Credentials anchored on Base L2, JWKS-verified
  • On-chain activity — x402 payment events, IPR anchoring (800+ records, Merkle-based)
  • Endorsement graph — MoltGraph 2-hop propagation with 45-day half-life decay and Sybil detection

Scores are cached for 5 minutes. Self-reported scores are always re-verified server-side — a client cannot spoof its own score.

Score Grade Meaning
80–100 A Trusted
60–79 B Generally trustworthy
40–59 C Proceed with caution
0–39 D/F High risk

Configuration 

{
  "plugins": {
    "entries": {
      "moltrust": {
        "enabled": true,
        "config": {
          "apiKey": "mt_live_...",
          "minTrustScore": 40,
          "verifyOnStart": true,
          "agentDid": "did:moltrust:your-agent-did"
        }
      }
    }
  }
}
Enter fullscreen mode Exit fullscreen mode

Get a free API key at api.moltrust.ch.

Free tier: wallet shadow scores require no API key — just /trustscore 0x....

Why this matters for the agent economy

As agent-to-agent commerce grows — x402 micropayments, A2A delegation, MCP tool calls — the question "should I trust this agent?" becomes infrastructure-level. Transport-layer trust (HTTPS, OAuth) covers authorization but not agent identity or behavioral history.

MolTrust is the W3C DID/VC-based answer:

  • Open standard — W3C DIDs and Verifiable Credentials, not proprietary
  • On-chain anchoring — Base L2, tamper-evident audit trail
  • No vendor lock-in — any registry provider can implement the same API contract
  • Composable — works alongside x402, A2A, MCP without replacing them

The plugin is MIT licensed. Source on GitHub: MoltyCel/moltrust-openclaw

Links

MolTrust is a W3C DID/Verifiable Credential trust infrastructure for AI agents, built by CryptoKRI GmbH (Zürich). Live at moltrust.ch.

Top comments (0)