Overview
AAP Solutions combines normalized public market data from supported perpetual venues with an agent-oriented strategy interface. The integration model separates market-data access, strategy evaluation and execution authorization so an AI agent can research opportunities without automatically receiving unlimited trading power.
Connect an agent
import { AAP } from "@aap-solutions/agent";
const agent = await AAP.connect({
key: process.env.AAP_API_KEY,
wallet: process.env.AGENT_WALLET,
permissions: [
"market.read",
"quote.create",
"trade.request"
],
maxNotionalUSD: 25000
});
The agent interface is designed around permissioned market-data access, quoting and execution requests.
Scan funding
const routes = await agent.funding.scan({
symbols: ["BTC", "ETH", "SOL"],
minNetApr: 10,
minOpenInterestUSD: 5_000_000,
venueTypes: ["dex", "cex"]
});
Create a route quote
const quote = await agent.quote(routes[0], {
notionalUSD: 10000,
includeFees: true,
includeSlippage: true,
includeFundingHistory: true
});
A quote should contain the long and short venue, executable prices, fees, estimated slippage, funding cadence, funding history summary and a risk score. Agents should make decisions on the quote, not on a raw headline APR.
Request execution
await agent.strategy.open(quote, {
mode: "delta-neutral",
maxDeltaBps: 25,
maxSlippageBps: 18,
timeInForce: "IOC"
});
Execution is intended to remain policy-bound. A production system should require explicit wallet authorization, venue-level permissions and per-strategy notional limits.
Risk policy
Funding arbitrage is not risk-free. An agent policy should account for basis divergence, asymmetric fills, liquidation thresholds, venue solvency, collateral movement, funding reversals, borrow constraints, API downtime and stale order-book data.
- Set maximum strategy notional.
- Set maximum tolerated delta during partial fills.
- Reject routes with insufficient depth.
- Require minimum collateral buffer on both legs.
- Define automatic exit conditions when funding edge decays.