Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.ameo.agiwithai.com/llms.txt

Use this file to discover all available pages before exploring further.

Autonomous agents already move real capital — rebalancing treasuries, routing grants, executing DEX trades. What they rarely provide is an independent proof trail: a way for operators, auditors, or the public to verify why a specific action happened without trusting a dashboard screenshot or a model’s post-hoc explanation. AMEO closes that gap. It is a full-stack runtime for verifiable cognition: an agent that observes on-chain and market state, compiles a structured plan, passes deterministic policy gates, executes on Mantle, and anchors both an on-chain decision fingerprint and an off-chain reasoning receipt.

​
The problem

Traditional “AI trading bots” optimize for speed and opacity. When something goes wrong — a bad swap, a policy bypass, a runaway loop — there is no canonical artifact to replay. Logs are mutable. Rationale is prose. On-chain txs show what moved, not why it was authorized. For protocol teams, researchers, and auditors, that opacity is the blocker for production adoption of autonomous finance.

​
What AMEO does

Every cognition cycle follows the same pipeline:
  1. Observe — snapshot treasury balances, gas, and market signals from Mantle RPC and configured feeds.
  2. Reason — compile a structured action plan (with LLM or deterministic fallback) and hash the rationale.
  3. Policy — evaluate hard predicates outside the model; failed plans never reach signing.
  4. Execute — broadcast policy-approved transactions via the Mantle DEX adapter.
  5. Prove — emit DecisionLogged on an ERC-8004-inspired identity contract and optionally anchor the full reasoning blob on 0G Storage.
The Narrative Console replays any recorded cycle across nine verification nodes — from observation through settlement.

​
How it is different

Typical agent stackAMEO
Model decides and executesModel proposes; policy engine authorizes
Logs in a databaseAppend-only JSONL event store + on-chain log
”Trust our UI”Explorer links + replay API + optional 0G receipt
Single LLM vendorProvider fallback chain with safe local_rules default
Every action can be traced from thought to trade with cryptographic commitments at each layer.

​
Who it is for

  • Protocol teams running treasury or agentic wallet programs on Mantle.
  • Researchers studying policy-bound autonomous agents and verifiable cognition.
  • Auditors and operators who need replay, not narration.
Ready to run it locally? Continue to the Quickstart.