Agent AQI
A trust scoreboard for on-chain agents
About
The problem it solves Autonomous on-chain agents can execute trades, manage liquidity, and automate DeFi strategies, but users have no reliable way to evaluate them. Today, choosing an agent means trusting: Marketing claims GitHub stars Twitter threads Vibes There is no standardized benchmark for: Execution reliability Slippage discipline Gas efficiency Route selection behavior On-chain verification Agent AQI introduces a trust scoreboard for autonomous DeFi agents. Users can: Race multiple agents against each other in real-time Inspect Uniswap route decisions and slippage policies Verify swap receipts via QuickNode Streams Compare performance metrics in a unified Agent Quality Index (AQI) Instead of guessing which agent to trust, users can rely on transparent, measurable performance data. Agent AQI acts as a credit score for autonomous agents in the emerging AI-driven DeFi economy. Use of AI tools and agents Agent AQI benchmarks three autonomous trading agents: **SafeGuard** Tight slippage tolerance Low hop count Rarely flagged Reliability-focused ** SpeedRunner** Higher slippage tolerance No hop limits Prioritizes execution latency Higher failure risk **GasOptimizer** Constrained routing Minimal gas usage Deadline risk under volatile conditions **Each agent:** Requests a live quote from the Uniswap Trading API Applies its own execution policy Simulates or executes the trade Emits structured metrics: Slippage (bps) Gas estimate Latency Success/failure flags QuickNode Streams acts as the primary source of truth, ingesting real swap receipts and updating AQI dynamically. Blockade Labs Skybox provides each agent with a persistent 360° spatial “home,” solving the “Homeless Agent” problem by anchoring agents in contextual digital environments. Together, these components form a complete trust layer for autonomous on-chain agents.
How it's made
Challenges we ran into **1. Uniswap API Was Not “Plug and Play”** On paper, calling the Uniswap Trading API is simple. In reality, I kept getting: “No quotes available” Even with valid token addresses. The issues turned out to be: Some token pairs had no liquidity on testnet Slippage tolerance was too strict for volatile routes Address validation was rejecting checksum addresses Chain ID mismatches silently broke requests I fixed this by: Cleaning up address validation logic Handling 404 responses properly instead of crashing the flow Surfacing route + hop count in the UI to debug what Uniswap was actually doing Adjusting slippage policies per agent This turned the Swap Simulator from something fragile into something demo-proof. **2. QuickNode Webhook Verification Broke Everything** The hardest bug was subtle. QuickNode sends webhooks with raw JSON bodies that must be verified using HMAC-SHA256. But Express parses JSON by default. If express.json() runs before express.raw(), the raw body is lost — and signature verification fails. It took debugging to realize middleware order was the issue. Fix: Mounted express.raw() before express.json() Used constant-time comparison for signature validation Built a live Streams monitor to prove verification works After that, the system could trust real on-chain data.