MarketTrace for AI agents

The agent feed is a read-only MCP server over the same cross-exchange microstructure that powers this site: 6 assets (BTC, ETH, SOL, BNB, XRP, DOGE) across Binance, Bybit, OKX and Hyperliquid. Facts and normalization, no verdicts — every metric carries its own coverage (venues, window depth, freshness), so an agent knows exactly how much to trust each number.

Connect

In Claude: Settings → Connectors → Add custom connector.
Paste the server URL: https://api.markettrace.ai/mcp
Authorize when prompted (email magic link — no API keys, nothing to copy).
Ask away. The feed is read-only; it can never trade or move funds.

https://api.markettrace.ai/mcp

Listed in the official MCP registry as ai.markettrace/agent-feed.

Open-source bridge & docs on GitHub

Tools

get_market_state — One normalized cross-exchange snapshot: funding + multi-year percentile, OI, volume, CVD, order-book imbalance, liquidations, basis, drivers.
get_funding_percentile — Current funding rate ranked against its own multi-year history (0–100) with the same-sign streak.
get_liquidations_recent — Cross-exchange liquidation totals for a window: USD totals, long/short split, event count.
get_ohlcv — Consolidated cross-exchange candles (5m…1d) for ATR, ranges and realized-volatility math.
get_conditional_outcomes — Measured history of forward returns after a stated condition (funding percentile / streak / sign / archived features) — base rates instead of folklore, honest history_silent when evidence is thin.
get_state_history — Time series of any numeric market-state field from the 15-minute archive — the trend view behind the snapshot.

Things to ask

“What's the market state for BTC — is positioning stretched?”
“What happened historically after funding above the 90th percentile?”
“How did open interest build over the last 3 days?”
“How much got liquidated on ETH in the last hour — longs or shorts?”

Honesty model

The feed reports what it can measure and says so when it can't: thin history answers with disclosed depth instead of made-up numbers, conditional outcomes go history_silent below the evidence floor, and every response self-declares freshness. Reports history, not predictions.