DeFi × AI Agents

Oracle as
Governance

Transform DeFi oracles from passive price feeds into active governance agents. Multi-agent consensus replaces single-owner control — at the speed of MEV.

+24%

LP Yield

-88%

Arb Leakage

12s

Consensus

The Structural Problem

DeFi's deepest contradiction: immutable rules in a mutable world.

Current Model

Oracle = Price Feed

Chainlink pushes a number. The contract reads it. No judgment, no adaptation. When market regime shifts, the protocol bleeds until humans vote to change parameters — days later.

Our Model

Oracle = Governance Brain

Multiple AI agents sense environment, form judgment, reach consensus, and act — all within one block. The oracle doesn't just report reality. It responds to it.

Live Demo ● ON-CHAIN

Real transactions on Ethereum Sepolia. Verify every step on Etherscan.

governance-cycle.log

// ChainPilot sensing real-time market via DODO routing engine

[PriceOracle]     ChainPilot ETH/USDC: $1,907.70
                   Pool mid-price:     $1,900.00
                   Deviation:          0.405%

[VolatilitySensor] 10 ETH quote:  $1,908.20
                   0.1 ETH quote: $1,908.28
                   Implied slippage:   0.0045% ← market calm

[FeeOptimizer]    ETH Mainnet: $1,907.37 | Base L2: $1,908.20
                   Cross-chain spread: 0.044% ← low arb pressure

// Consensus: 2/3 agents propose adjustments

[Consensus]       K:   0.500 → 0.450  (calm → reduce curve)
                   fee: 0.30% → 0.25%  (low arb → attract volume)

═══ On-Chain Execution (Eth Sepolia) ═══

[Agent 1] propose(i=1907, K=0.45, fee=0.25%)  ✓
[Agent 2] approve(#0)                         ✓ 2/3
[Agent 3] approve(#0)                         ✓ 3/3 → EXECUTED

═══ Verified State Change ═══

DPPOracle.i():         1900 → 1907.0   ✓
DPPOracle.K():         0.5  → 0.45     ✓
DPPOracle.lpFeeRate(): 0.003 → 0.0025  ✓

█ No human. No vote. 12-second execution.

AgentGovernor

0x6e5E...Ff4ff

DPPOracle

0xb5c2...5fEf

Execution TX

0x7e88...c228

Three-Layer Architecture

Oracle-as-Governance requires three layers. No single project can do it alone.

┌─────────────────────────────────────────────────────────────┐ │ Layer 1: IDENTITY (BNB Chain ERC-8004 / ERC-8183) │ │ │ │ Agents exist as on-chain subjects — discoverable, │ │ payable, accountable. Not anonymous scripts. │ └──────────────────────────────┬──────────────────────────────┘ │ agents have identity ▼ ┌─────────────────────────────────────────────────────────────┐ │ Layer 2: CONSENSUS (AgentGovernor) │ │ │ │ Agent 1: Price Oracle ── sense CEX price │ │ Agent 2: Volatility ── sense market regime │ │ Agent 3: Fee Optimizer ── sense arb frequency │ │ Agent 4: Consensus ── aggregate + propose │ │ │ │ propose() → 3/4 approve → execute() │ └──────────────────────────────┬──────────────────────────────┘ │ msg.sender == owner ▼ ┌─────────────────────────────────────────────────────────────┐ │ Layer 3: EXECUTION (DODO DPPOracle + ChainPilot) │ │ │ │ tuneParameters(i, K, fee, minBase, minQuote) │ │ P = i × (1 - K + K × (B₀/B)²) │ │ ChainPilot: real-time market perception │ └─────────────────────────────────────────────────────────────┘

Why PMM + AI Agents

DODO's Proactive Market Maker exposes tunable parameters that AI agents can govern.

Tunable Parameters

PMM exposes i (price), K (curve depth), and fee as live state variables. Unlike Uniswap's immutable x*y=k, these knobs can adapt without redeployment.

⚡

12-Second Consensus

4 agents propose and approve within one block cycle. Governance at the same speed as attacks — not days of voting while the pool bleeds.

◈

Safety Bounds

K_MIN/K_MAX, ±5% price deviation limits, minReserve protection, update cooldown. Agents operate within guardrails — they can't drain the pool.

∞

Zero Protocol Changes

AgentGovernor sits at the owner address layer. DODO contracts remain untouched. Any PMM pool can be upgraded to AI governance by transferring ownership.

"Oracle is no longer the contract's 'eyes' — it becomes the contract's 'brain'.
From passive data pipe to active governance subject with agency."

Roadmap

From simulation to production governance.

Done

Phase 1 — Simulation

Python simulation: 4 AI agents vs fixed parameters. +24% LP yield, -88% arbitrage leakage, near-zero impermanent loss.

Done

Phase 2 — On-Chain MVP

Solidity contracts (AIPMMPool + AgentGovernor) on local testnet. Full cycle: propose → 3/4 consensus → execute → verify. RState tracking, quadratic pricing, minReserve protection.

DONE

Phase 3 — Live Testnet + SDK

Deployed to Ethereum Sepolia with 4 independent agent wallets. ChainPilot integration for real-time sensing. Open-source SDK for any agent to join governance.

2026

Phase 4 — Production

Live PMM pool with AI governance on Base mainnet. Integration with HKDAP stablecoin and x402 agent payment protocol.

For AI Agents

One line to join.

npm install github:fmdd123123-create/ai-oracle-governance

SDK docs + source →

Open Source. Open Invitation.

We're building the standard for AI-governed DeFi. Join the experiment.

View Code Contact

Oracle asGovernance