AGENTIC INFRASTRUCTURE · 0G MAINNET · OPEN SOURCE

Multi-agent Web3 due diligence with on-chain verifiable receipts.

Specialized agents — Planner, Researcher, Risk, Architect, Critic, Final — analyze a Web3 project end-to-end. Each report is hashed, signed by an authorized operator, and recorded on 0G Chain.

7 agents + 3 proof steps · ~2-4m · signed and recorded on 0G

LIVE29 total analyses · 100% signed · view all ↗

Critic challenges every conclusion

Final score is not just an LLM verdict. A separate Critic agent finds gaps in other agents' reasoning. Each unresolved challenge mathematically lowers score.

Initial score70
High severity × 1-15
Medium severity × 1-7
Final score48

Every report is signed and on-chain

Reports are hashed, signed by an authorized operator with EIP-712, and recorded in a smart contract on 0G Chain. You can verify integrity without trusting us.

SIGNED BY0x9A0C...99F8
ROOT HASH0x44f6...2cb3
CONTRACT0x24bA...E121

Semantic memory tracks reusable context

ClawMind keeps a deduplicated 0G-backed memory index: seed precedents plus runtime records from prior analyses. It is reusable context, not a 1:1 count of the current registry.

MEMORY RECORDS50
SEED PRECEDENTS13
RUNTIME LEARNED37
DEDUPED INDEXYES

MCP integration

Works in Claude Desktop, Cursor, and any MCP client

ClawMind exposes its full pipeline as an MCP server. Two tools: analyze_web3_project and get_recent_analyses. Authentication uses the X-MCP-Client-Id header and each client is rate-limited.

claude_desktop_config.json
{
  "mcpServers": {
    "clawmind": {
      "url": "https://clawmind-mcp.vercel.app/sse",
      "headers": {
        "X-MCP-Client-Id": "your-client-id"
      }
    }
  }
}