Autonomous Optimization Engine (AOE)
AlgoVault’s closed-loop learning system. The Autonomous Optimization Engine continuously tunes AlgoVault’s composite trade-call model from published trade outcomes. Each call adds to the dataset; the dataset improves the model; improvements flow back through every API call. Result: PFE win rate has improved since launch (current: 90.4%).
Like an LLM that gets sharper with every interaction, AOE is the substrate that makes AlgoVault “The Trading Model API” rather than a static rule set. See /how-it-works and /track-record.
Composite verdict
AlgoVault’s primary output unit. A single BUY/SELL/HOLD with confidence, regime label, and contributing factors. Multi-factor quant weighting across price, funding rate, open interest, and volume. Exposed via get_trade_call. See /track-record.
Cross-venue intelligence
Aggregating market data and detecting opportunities across multiple exchanges (Hyperliquid, Binance, Bybit, OKX, Bitget) rather than a single venue. Enables funding-rate arbitrage detection and venue-aware regime classification.
Data flywheel
Every published trade call generates verifiable outcome data. That data feeds the composite weighting tuner, which improves signal quality, attracts more usage, and generates more data. Visible at /track-record.
Funding arbitrage
A cross-venue trading strategy that exploits differences in perpetual-swap funding rates between exchanges. AlgoVault’s scan_funding_arb MCP tool returns ranked opportunities by annualized basis-point spread across Hyperliquid, Binance, and Bybit. See /skills for ready-made agent recipes.
HOLD rate
The percentage of trade-call queries that return a HOLD verdict (no actionable BUY or SELL signal). Currently 98.8% on a rolling 30-day window. HOLDs are always free at every pricing tier; selectivity is AlgoVault’s primary upsell mechanic.
MCP (Model Context Protocol)
An open protocol that lets AI agents call external tools and resources over Streamable HTTP or stdio. AlgoVault is an MCP server (the producer of trade calls); agents like Claude Desktop, Cursor, and Cline are MCP clients (the consumers). See /docs for connection examples.
Merkle anchor
A cryptographic commitment that proves a set of trade calls existed at a specific block. AlgoVault batches signals into Merkle trees and publishes the root to a Base L2 smart contract (0x6485…bf81). The /verify page reconstructs the proof tree for any individual call.
PFE win rate
Path-Forward Equity win rate. The percentage of AlgoVault’s BUY/SELL trade calls that closed positive on the trader’s intended path forward. Currently 90.2% across 79,527 verified calls on a rolling 30-day window. Refetched live from /api/performance-public.
Platform ecosystem
The network of MCP-compatible clients (Claude Desktop, Cursor, Cline, Smithery, custom agents) and downstream skills and recipes that consume AlgoVault output. Each new client connector strengthens the integration surface. See /skills.
Published track record
Live, refetchable performance numbers exposed via /api/performance-public, with every signal Merkle-anchored on Base L2 and verifiable at /verify. Distinct from claimed-but-unverifiable backtests.
Regime classifier
A model that labels the current market state as trending, ranging, or volatile based on price action, funding sentiment, and cross-venue flow. Exposed via AlgoVault’s get_market_regime MCP tool, which also returns strategy and sizing suggestions appropriate to the labeled regime.
Signal MCP
A category of MCP server that produces analytical signals (composite trade calls, regime classification, arbitrage spreads) for AI-agent consumption. Distinct from execution-MCP (places orders) and data-MCP (returns raw OHLC). AlgoVault is the canonical reference implementation of a signal MCP.
Suite lock-in
The ergonomic advantage when an AI agent depends on multiple AlgoVault MCP tools (get_trade_call plus scan_funding_arb plus get_market_regime). Switching costs grow with each integration; consistency of the _algovault metadata block makes cross-tool reasoning easier.
Trade call
AlgoVault’s primary unit of output: a structured tuple {call, confidence, regime, factors} for a given (asset, timeframe, exchange) triple. Issued by the get_trade_call MCP tool. Each trade call is hashed into the next Merkle batch published on Base L2.
x402 micropayments
An HTTP 402 protocol for pay-per-API-call settlement using USDC on Base L2. AlgoVault accepts x402 payments as an alternative to subscription tiers, settling per-call without account creation. Note: x402 is revenue, not verification — distinct from Merkle-anchor track record.