MarketTrace Crypto MCP server

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

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

Data: funding rates, open interest, cumulative volume delta (CVD), order-book depth, liquidations, OHLCV candles.

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).

Claude Code:

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

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

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

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

Local stdio bridge (this repo)

mcpserver.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 MARKETTRACEBEARER 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, generatedat, 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 getmarket_state).

python3 mcp_server.py            # Python 3.9+, no dependencies

Or with Docker:

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?"

Terms

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

Shortened. The full README is on GitHub.