Skip to main content
AMEO splits proof across two layers optimized for cost, availability, and finality.

​
Layer 1 — 0G Storage (reasoning blob)

The full planner output — rationale text, structured plan JSON, model metadata — can be uploaded to 0G Storage. The worker receives a root hash verifiable via the 0G indexer.
  • Pros: Large payloads, cheap retention, public read via indexer URL.
  • Cons: Off-chain; requires 0G configuration (ZERO_G_RPC_URL, ZERO_G_PRIVATE_KEY, CLI path).
When not configured, replay shows an honest empty state — never a fabricated hash.

​
Layer 2 — Mantle (commitment + decision log)

On-chain, MantleAgentIdentity.logDecision stores:
  • rationaleHash — commitment to the rationale
  • dataHash — often the 0G root or auxiliary blob hash
  • actionType, pnl1e18, metadataUri
The settlement transaction hash proves execution; DecisionLogged proves authorization tied to agent identity.

​
Why the split matters

An auditor can validate commitments on-chain first, then fetch the blob off-chain if needed — standard commit–reveal for autonomous agents. Configure 0G in .env — see Quickstart. Contract details: Agent identity.