ai.markettrace/agent-feed MCP Server

The ai.markettrace/agent-feed MCP server provides read-only crypto perpetual-futures microstructure data intended for AI agents. It delivers a normalized cross-exchange market state covering funding rates, open interest, liquidations, order book, and CVD. Coverage and freshness are self-declared per metric, and the agent performs interpretation.

🛠️ Key Features

Read-only access to crypto perps microstructure
Normalized cross-exchange market state
Metrics include funding rates, open interest, liquidations, order book, and CVD
Self-declared coverage and freshness on every metric
Hosted MCP endpoint: https://api.markettrace.ai/mcp (Streamable HTTP, OAuth; no API keys)

🚀 Use Cases

Supplying AI agents with perpetual-futures market data
Building agents that consume order book and orderflow-derived signals (e.g., CVD)
Monitoring perps risk factors like funding rates and liquidations

⚡ Developer Benefits

Streaming HTTP transport for market data delivery
OAuth-based access without API keys
Server is listed in the official MCP registry as ai.markettrace/agent-feed

⚠️ Limitations

Designed for read-only market data; no trading actions are described
Only the documented metric set (funding rates, open interest, liquidations, order book, CVD) is supported

MarketTrace agent-feed

Read-only crypto perps microstructure for AI agents — normalized cross-exchange market state with self-declared coverage and freshness on every metric. Facts and normalization, no verdicts: the agent interprets.

Hosted MCP: https://api.markettrace.ai/mcp (Streamable HTTP, OAuth — no API keys)
Official registry: ai.markettrace/agent-feed
Docs: https://markettrace.ai/agents

This repo is the front door — connection configs, the interface contract, and a thin stdio bridge. The data pipeline itself (4-venue ingest, archives, normalization) is not open source.

What it serves

9 assets (BTC, ETH, SOL, BNB, XRP, DOGE, HYPE, ZEC, ENA) across Binance, Bybit, OKX, Hyperliquid:

Tool	What it answers
`get_market_state`	One normalized snapshot: funding + its trailing 2-year percentile, OI, volume, CVD, order-book imbalance, liquidations, basis, drivers. "Is ETH positioning stretched?"
`get_funding_percentile`	Current funding ranked against its own trailing 2-year window (0–100) + same-sign streak.
`get_liquidations_recent`	Cross-exchange liquidation notional estimates for a window: USD, long/short split.
`get_ohlcv`	Consolidated cross-exchange candles (5m…1d) with per-candle delta (taker buy − taker sell), for ATR, ranges, CVD and RV math.
`get_conditional_outcomes`	Measured forward-return history after a stated condition — base rates instead of folklore. "What happened historically after funding above the 90th percentile?"
`get_state_history`	Time series of any numeric state field from the 15-minute archive — the trend view behind the snapshot.
`get_volume_profile`	Volume-profile levels per UTC day from the consolidated tape: POC, value area high/low, value-area width and its rank, a multi-day composite and naked POCs. "Is price inside yesterday's value area?"
`get_big_trades`	Large aggressive orders (fills sharing venue, side and timestamp summed into one trade): per-side totals plus the biggest prints with venue, price and USD size. "Were the whale market orders buying or selling?"
`get_footprint_events`	Order-book wall events from the 1-minute footprint: absorbed and pulled walls with peak, executed and closing size in USD, plus thin-book minutes. "Were bid walls pulled before this drop?"
`get_stacked_imbalances`	Stacked footprint imbalances from the consolidated tape: diagonal buy and sell runs per candle (1m…1h) with price band and USD size. "Where did aggressive buyers stack up on BTC this morning?"

Data: funding rates, open interest, cumulative volume delta (CVD), order-book depth, liquidations, OHLCV candles with per-candle delta, volume-profile levels, large aggressive orders, order-book wall events, stacked footprint imbalances.

Honesty model: every metric carries a coverage entry (venues, window depth, freshness); thin history answers with disclosed depth instead of made-up numbers; conditional outcomes go history_silent below the evidence floor; every response self-declares its age. Reports history, not predictions.

Connect

Claude (web/desktop): Settings → Connectors → Add custom connector → https://api.markettrace.ai/mcp → authorize (email magic link).

Every client signs in the same way: it opens a browser, you enter your email, and a sign-in link arrives. Open it on the same device and in the same browser where you started.

Claude Code (adding the server does not sign you in, so run both):

bash

claude mcp add --transport http markettrace https://api.markettrace.ai/mcp
claude mcp login markettrace

Codex:

bash

codex mcp add markettrace --url https://api.markettrace.ai/mcp
codex mcp login markettrace

Cursor: add this to ~/.cursor/mcp.json (or use Add to Cursor on https://markettrace.ai/agents) and sign in when Cursor asks:

json

{
  "mcpServers": {
    "markettrace": { "url": "https://api.markettrace.ai/mcp" }
  }
}

Stdio-only clients (via the standard OAuth-capable bridge):

bash

npx -y mcp-remote https://api.markettrace.ai/mcp

More client configs in examples/mcp-configs.md.

Local stdio bridge (this repo)

mcp_server.py is a zero-dependency stdio bridge: it starts and answers introspection (initialize, tools/list) with no credentials — the bundled tools.json is a snapshot of the hosted server's own contract. Tool calls are proxied to the hosted endpoint when MARKETTRACE_BEARER is set; without it they return a pointer to the hosted OAuth endpoint instead of data. It holds no methodology — just a client.

Refresh the contract: tools.json is a {version, generated_at, tools} snapshot of the live server's tools/list — regenerate it by capturing that response and stamping the current contract version (mirrors feed.version in get_market_state).

bash

python3 mcp_server.py            # Python 3.9+, no dependencies

Or with Docker:

bash

docker build -t markettrace-bridge . && docker run -i markettrace-bridge

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?"
"Where are the key volume levels on BTC — any untested POCs nearby?"
"Did a large resting order get absorbed on SOL in the last hour?"

Terms

Informational market data only — not financial advice. Privacy Policy · Terms of Service · Contact: support@markettrace.ai

The bridge in this repo is MIT-licensed (LICENSE); the hosted service is governed by the Terms above.

from	string \| null	optional	Window start: RFC3339, `YYYY-MM-DD` (UTC midnight) or epoch ms. Default: `to` minus one hour. No older than 31 days.
limit	integer \| null	optional	Trades listed, 1..=150 (default 50); the largest are kept.
min_usd	integer \| null	optional	Smallest notional that counts, whole USD, at least 50000. Default: the asset's chart threshold (echoed in the response). Read as any JSON value and checked in [`big_trades_query`], so a fractional, negative or non-numeric value is an invalid_params NAMING the parameter (a typed `u64` would fail inside rmcp's deserializer with serde's "expected u64" and no field name). The schema says integer.
symbol	string	required	Asset symbol: BTC, ETH, SOL, BNB, XRP, DOGE, HYPE, ZEC or ENA (case-insensitive; a `USDT` suffix is accepted).
to	string \| null	optional	Window end (exclusive), same formats. Default: now. At most 12 h after `from`.

archive_feature	string \| null	optional	v2: condition on ONE archived state feature (15-min archive): `obi_skew` (−1..+1), `oi_chg_1h_pct` (%), `taker_buy_ratio` (0..1), `basis_bps`. ANDs with the funding fields; the match window becomes the INTERSECTION of the constrained series' spans. Needs at least one archive threshold below.
archive_gte	number \| null	optional	Match hours where the archive feature's RAW value ≥ this (raw thresholds have no sample floor — they exist from the first point).
archive_lte	number \| null	optional	Match hours where the archive feature's RAW value ≤ this.
archive_pct_gte	number \| null	optional	Match hours where the feature's AS-OF percentile ≥ this (0..=100; the rank exists only once its series has warmup_samples points).
archive_pct_lte	number \| null	optional	Match hours where the feature's as-of percentile ≤ this (0..=100).
funding_pct_gte	number \| null	optional	Match hours where the AS-OF funding percentile was ≥ this (0..=100). "As-of" = the rank the live endpoint reported at that moment — computed only from data available then, no look-ahead.
funding_pct_lte	number \| null	optional	Match hours where the as-of funding percentile was ≤ this (0..=100).
funding_sign	string \| null	optional	Funding sign filter: `positive` (longs pay shorts) or `negative`.
horizons	array \| null	optional	Forward horizons as `<hours>h` labels, each 1h..=168h (default `["4h","24h","72h"]`).
streak_h_gte	integer \| null	optional	Match hours where funding had kept its sign for at least this many hours.
symbol	string	required	Asset symbol: BTC, ETH, SOL, BNB, XRP, DOGE, HYPE, ZEC or ENA (case-insensitive; a `USDT` suffix is accepted).

from	string \| null	optional	Window start: RFC3339 (`2026-10-12T09:10:00Z`), `YYYY-MM-DD` (UTC midnight) or epoch milliseconds. Default: `to` minus one hour.
kinds	array \| null	optional	Event kinds to return: `absorbed` and/or `pulled` (default both).
limit	integer \| null	optional	Events returned, 1..=150 (default 50). When the window holds more, the NEWEST are kept and `truncated` is true; `counts` still cover the window.
min_usd	number \| null	optional	Smallest `peak_usd` that counts, USD, >= 0 (default 0 = every event).
symbol	string	required	Asset symbol: BTC, ETH, SOL, BNB, XRP, DOGE, HYPE, ZEC or ENA (case-insensitive; a `USDT` suffix is accepted).
to	string \| null	optional	Window end (exclusive), same formats. Default: now. At most 24 h after `from` — longer is an error, not a clamp.

symbol	string	required	Asset symbol: BTC, ETH, SOL, BNB, XRP, DOGE, HYPE, ZEC or ENA (case-insensitive; a `USDT` suffix is accepted).
window_s	integer \| null	optional	Look-back window in seconds, 1..=86400 (default 3600). Values beyond 24 h are rejected, not clamped.

interval	string \| null	optional	Candle interval: 5m, 15m, 1h, 4h or 1d (default 1h).
limit	integer \| null	optional	Number of candles, newest last: 1..=200 (default 100).
symbol	string	required	Asset symbol: BTC, ETH, SOL, BNB, XRP, DOGE, HYPE, ZEC or ENA (case-insensitive; a `USDT` suffix is accepted).

MarketTrace agent-feed

ai.markettrace/agent-feed MCP Server

🛠️ Key Features

🚀 Use Cases

⚡ Developer Benefits

⚠️ Limitations

Topics

get_big_trades

get_conditional_outcomes

get_footprint_events

get_funding_percentile

get_liquidations_recent

get_market_state

get_ohlcv

get_stacked_imbalances

get_state_history

get_volume_profile

MarketTrace agent-feed

What it serves

Connect

Local stdio bridge (this repo)

Things to ask

Terms

Install

Remote endpoint

Related servers

fields	array	required	Dotted numeric paths of the market state, 1..=16 per request — e.g. `funding.percentile`, `oi.usd`, `obi.skew`, `cvd.delta_usd`, `basis_bps`, `price.last`, `depth.bid_usd.10bps`, `venues.binance.oi_usd`. Typos get an error naming the offender.
from	string \| null	optional	Range start: RFC3339 (`2026-07-03T08:00:00Z`), `YYYY-MM-DD` (UTC midnight), or epoch milliseconds. Default: the whole archive.
max_points	integer \| null	optional	Points returned after downsampling, 1..=1000 (default 200). The response's `stride` says how many archived rows each point skips.
symbol	string	required	Asset symbol: BTC, ETH, SOL, BNB, XRP, DOGE, HYPE, ZEC or ENA (case-insensitive; a `USDT` suffix is accepted).
to	string \| null	optional	Range end, same formats. Default: the whole archive.

days	integer \| null	optional	Complete UTC days to return, 1..=30 (default 7). Out of range is an error, not a clamp.
symbol	string	required	Asset symbol: BTC, ETH, SOL, BNB, XRP, DOGE, HYPE, ZEC or ENA (case-insensitive; a `USDT` suffix is accepted).