Skip to main content

​
ClawMind

Multi-Agent Web3 Cognitive Backbone
ClawMind runs production LLM agent inference through the 0G Compute network using deepseek/deepseek-chat-v3-0324, and anchors reports on 0G Chain (Chain 16661).

​
The Problem

Web3 investors, DAO contributors, and protocol teams routinely make high-stakes decisions — approving grants, listing tokens, deploying integrations — based on scattered notes, subjective risk calls, and reports that nobody can reproduce later. Three things break this process:
  • No audit trail. Who analyzed what, when, and why? There’s no record.
  • No adversarial challenge. A single analyst’s blind spots become the report’s blind spots.
  • No integrity guarantee. A report stored in Notion or Google Docs can be quietly edited after the fact.

​
The Solution

ClawMind turns due diligence into a structured, verifiable, adversarial pipeline. A task enters the system and flows through 8 specialized agents — each with a distinct role. The Critic agent challenges the other agents’ conclusions before a final report is produced. That report is then stored on 0G Storage and its hash is anchored on 0G Chain, creating a tamper-evident receipt that anyone can verify at any time.
Every run produces:
  • A structured due diligence report with a score and recommendation (GO / INVESTIGATE_MORE / NO_GO)
  • A 0g:// storage URI pointing to the full report JSON
  • An on-chain transaction with the report hash, score, and an EIP-712 operator signature
  • A memory index entry that improves future analyses on similar projects

​
Key Differentiators

Adversarial by design

The Critic agent actively challenges weak assumptions from the Researcher, Risk, and Architect agents before the Final agent produces a verdict. Unresolved challenges reduce the score by severity.

On-chain integrity

Report hashes are recorded in AnalysisRegistry.sol on 0G Chain with an EIP-712 operator signature. The result cannot be altered without breaking the on-chain proof.

Persistent memory

ClawMind stores a semantic memory index on 0G Storage. Each run retrieves relevant context from past analyses — improving accuracy on similar projects over time.

OpenClaw compatible

ClawMind exposes an OpenClaw manifest describing the 8-step pipeline, artifacts, and security policies — making it composable with other agent systems in the 0G ecosystem.

​
How Scoring Works

Every analysis produces a score from 0 to 100 and maps to one of three recommendations. The Critic applies score penalties based on the severity of unresolved challenges:

​
0G Integration at a Glance

ClawMind uses four 0G components in production:

​
Safety Notice

ClawMind is a due diligence aid, not a formal security audit or exploit detector. The LLM pipeline produces structured analysis; on-chain receipts make results verifiable. Any deployment touching custody, signing keys, protocol upgrades, or automated execution must still require deterministic policy gates and human approval outside the model.