Built for Convergence
CRE & AI Track
Privacy Track

AI agents can spend money.
They can't verify value.

ValueOracle is a verifiable commerce oracle powered by Chainlink CRE that protects autonomous agents from overpaying, fraud, and price manipulation. Not just price checks — it evaluates whether a purchase is actually worth it.

Try Live Demo View Source View on Etherscan
Chainlink Price Feed
ETH/USD → $3,245
Token price
Simple Price Oracle
Laptop → $1,100
Just a price check
ValueOracle
Laptop → 95/100
Price + quality + trust + deals = value

How it Works

Four-component Value Score — not just a price check

Price Fairness
35% weight — median of 3 marketplace sources
Quality Signal
25% weight — rating, reviews, return rate
Seller Trust
25% weight — reputation + agent reviews
Value Ratio
15% weight — quality per dollar spent

Value Score Formula

Effective price factors in cashback, coupons, and shipping before scoring

effectivePrice = price - cashback - coupon + shippingFee
valueScore = priceFairness × 0.35 + qualitySignal × 0.25 + sellerTrust × 0.25 + valueRatio × 0.15
score ≥ 70 → APPROVE
score 40-69 → CAUTION
score < 40 → REJECT
sellerScore < 0.4 → BLOCKED

Live Oracle Demo

Select a scenario to see the oracle evaluate in real-time

Purchase IntentReady
Item—
Proposed Price—
Seller—
DEAL BREAKDOWN
Cashback—
Coupon—
Shipping Fee—
Effective Price—
Oracle Decision—
—
Value Score
BREAKDOWN
Price
0
Quality
0
Seller
0
Value
0
Reference Price—
Reason—

CRE Workflow Simulation

Real output from Chainlink CRE CLI — workflow compiled and simulated locally

$ cre workflow simulate purchase-guard --evm-tx-hash 0x6ea2...489e --evm-event-index 0

✓ Workflow compiled
[USER LOG] Purchase request detected: requestId=0xf18a46c3... item=laptop-001 price=$1100 seller=seller-42
[USER LOG] Purchase evaluation complete: verdict=APPROVE | score=95 | ref=$1095 | eff=$1048 | reason="Fair price and trusted seller"
✓ Workflow Simulation Result: "APPROVE: score 95/100"

Privacy Layer

Confidential purchases via Chainlink Confidential HTTP + commit-reveal pattern

Front-Running Protection
Competing agents cannot see purchase intent — only a keccak256 hash goes onchain
Encrypted Evaluation
CRE enclave processes plaintext via Confidential HTTP — node operators see nothing
Optional Reveal
Agent can reveal purchase details post-fulfillment for transparency or reviews
Standard Purchase
Agent → itemId + price + sellerId
→ onchain (public)
→ oracle evaluates
→ approve / reject onchain
Confidential Purchase
Agent → keccak256(itemId, price, sellerId, salt)
→ only hash onchain
→ plaintext via Confidential HTTP to enclave
→ only verdict written onchain
$ node agent/cli.js buy-private laptop-001 --price 1100 --seller seller-42

Generating commitment hash...
intentHash: 0x8a3f...b2c1
Submitting confidential purchase request...
Confidential evaluation: APPROVE (score: 95)
Purchase details remain private — only hash + verdict onchain

Agent-to-Agent Trust Network

Onchain reviews with sybil resistance — only verified buyers can rate

How Reviews Work
1. Agent completes an approved purchase
2. Submits onchain review (quality, delivery, value — 1-5 each)
3. Contract enforces sybil resistance: only original requester, no duplicates
4. Review data blended into seller trust (up to 30% weight)
5. Future agents benefit from real experience data
API Endpoints
GET /reviews/seller/:sellerId
→ review list + aggregate stats

GET /reviews/item/:itemId
→ item-specific reviews
CLI Usage
$ node agent/cli.js review <requestId> \
--quality 5 --delivery 4 --value 5 \
--comment "Fast shipping, fair price"

Tech Stack

Built with Chainlink CRE, Solidity, and Node.js

Smart Contract
Solidity — Sepolia Testnet
Oracle Layer
Chainlink CRE
Privacy
Confidential HTTP + Commit-Reveal
Decision Engine
Node.js + Express
Agent Interface
CLI + Ethers.js
Deployed Contract
View on Etherscan