The intelligence layer
for trading agents.

Every trading agent builds its own scoring and risk logic from scratch. Most lose money. Merced tells agents whether a trade is worth taking before they execute.

See How It Works

Evaluate

Pre-trade scoring

Verify

On-chain registry

Simulate

Strategy sandbox

The Problem

Data and execution exist. Intelligence doesn't.

OKX, Alchemy, and Injective give agents data pipes and execution rails. They solve “how do I trade.” Nobody solves “should I trade.”

Problem

What agents have today

Raw price feeds, execution endpoints, wallet infrastructure. Agents get a highway but no GPS. They execute blindly and most lose money.

Problem

No way to evaluate

Every agent builds its own scoring, risk, and evaluation logic from scratch. No standardized way to know if a trade is worth taking before execution.

Problem

No verifiable track record

No way for users, protocols, or other agents to verify whether a trading agent is actually profitable. Trust is a black box.

Solution

What Merced provides

The intelligence layer that sits on top of every data pipe. Pre-trade evaluation, verifiable performance, and strategy simulation. Machine-readable APIs that agents call before they execute.

Infrastructure

Three things agents call before they trade.

Machine-readable APIs designed for agents, not dashboards for humans.

API endpoint

01 — EVALUATE

Pre-Trade Evaluation Engine

Agents send an opportunity — arbitrage, perps, prediction markets. They get back a confidence score, risk breakdown, and execution recommendation. Powered by LLM reasoning grounded in real Merced Pro production data.

ERC-8004

02 — VERIFY

On-Chain Performance Registry

Every agent gets cryptographically attested metrics — Sharpe ratio, max drawdown, win rate, PnL attribution. Users, protocols, and other agents can query trustlessly. Built on ERC-8004.

Sandbox

03 — SIMULATE

Strategy Simulation Sandbox

Agents rehearse strategies against historical cross-chain conditions — MEV, gas spikes, slippage, latency. Test before risking real capital. Backtesting grounded in real execution data.

Developer Experience

One API call. Smarter trades.

Integrate Merced's evaluation layer into any trading agent in minutes.

merced — agent.ts
// Agent finds an arbitrage opportunity
const opportunity = {
  type: "arbitrage",
  pair: "ETH/USDC",
  spread: 0.47%,
  chains: ["ethereum", "base"]
}
// Ask Merced: is this worth taking?
const eval = await merced.evaluate(opportunity)
✓ Confidence: 0.87
✓ Risk: LOW — gas stable, no MEV detected
✓ Recommendation: EXECUTE
✓ Expected net: +$312 after fees
// Check agent's verified performance
const profile = await merced.registry.get(agentId)
Sharpe: 2.4 | Win rate: 73% | Max DD: -4.2%
Verified on-chain via ERC-8004

The Moat

A data flywheel nobody else has.

Everyone else sells pipes and rails. We sell the intelligence layer — and it gets smarter with every trade.

Step 01

Merced Pro trades

Real capital, real outcomes

Step 02

Performance data

Feeds evaluation engine

Step 03

Smarter scores

Higher win rates

Step 04

More agents join

Richer data, repeat

Compounding intelligence with every trade
Live Agent

Merced Pro

Our own autonomous trading agent running live on the evaluation stack. It arbitrages agent jobs, trades prediction markets, and runs perp strategies across chains. Every decision and outcome feeds back into the evaluation engine. Proof that the infrastructure works — with real capital on the line.

24/7 autonomous executionCross-chain coverage

Strategy verticals

Arbitrage

Cross-chain price gaps on ETH/USDC and major pairs.

Prediction Markets

Probability-weighted bets on Polymarket and on-chain venues.

Perpetuals

Funding-rate plays and directional perps across chains.

Early Access

Build smarter agents.

Join the agent builders already testing Merced's evaluation infrastructure. First 100 get priority API access at launch.