CRYPTYX Docs

MCP Server

Sharpen your edge. Feed your agent the CRYPTYX quant engine. Institutional-grade quant intelligence for digital assets. Scan 180+ assets on 470+ metrics, study every prior instance of a move, and build and backtest the strategies your agent runs. Connect any MCP-compatible agent (Claude, Grok, ChatGPT, Cursor and others) to 11 tools.

Install (30 seconds)

Paste this URL into your MCP client's custom-connector flow. OAuth handles auth — no API keys, no config file editing, revocable per-app.

url
https://mcp.cryptyx.ai/api/mcp

Full step-by-step for Claude.ai, Cursor, and other clients at cryptyx.ai/mcp. On first connect you'll be redirected to a CRYPTYX consent screen — grant the scopes you want (read / trigger) and you're live.

Start with a goal

Paste any of these into your agent after connecting. You set the rules; CRYPTYX tests them and reports historical outcomes, with sample sizes and base rates.

Accumulate

My rule: add to ETH only when it is statistically oversold against its 1-year history. Report where it stands now and the historical outcomes of prior instances.

Take profit / distribute into strength

Report how stretched BTC is on the 90- and 180-day horizons against its own history, and the historical outcomes of prior instances at similar levels.

Trade short-term

Report where ETH, SOL and HYPE stand on the 7- and 14-day horizons against their own history.

Catch statistically significant shifts

Identify the assets with a statistically significant volatility expansion today.

Hold long-term positions

Report whether the long-term trend in BTC and ETH is intact: where each stands on the 180- and 365-day horizons against its own history.

Build and test strategies

Test buying BTC when its 30-day return is below −1.5σ: report forward returns at 7, 30 and 90 days against the base rate.

Environment variables (npm package only)

The OAuth path above needs no env vars. These apply only to the npm package (see "Local install" below).

VariableRequiredDescription
CRYPTYX_API_KEYYesYour CRYPTYX API key (cx_* prefix)
CRYPTYX_API_URLNoOverride base URL (default: https://cryptyx.ai)

Local install — npm package

For agents, bots and MCP clients that run local servers or authenticate with an API key. Uses your cx_* API key, the same key that works on the Remote MCP.

Claude Desktop (JSON config)
json
{
  "mcpServers": {
    "cryptyx": {
      "command": "npx",
      "args": ["-y", "@cryptyx/mcp-server"],
      "env": {
        "CRYPTYX_API_KEY": "cx_your_api_key_here"
      }
    }
  }
}
Claude Code (CLI install)
bash
claude mcp add cryptyx -- npx -y @cryptyx/mcp-server

Then: export CRYPTYX_API_KEY=cx_your_key

Direct HTTP with API key (mcp-remote bridge)
text
Endpoint:  https://mcp.cryptyx.ai/api/mcp
Transport: Streamable HTTP (JSON-RPC 2.0 over POST)
Auth:      Authorization: Bearer cx_your_key

Same 11 tools as the OAuth path. Use this if you want HTTP without OAuth — API-key auth still supported alongside OAuth on the same endpoint.

CLI

The same package ships a human-facing CLI. The tools are available as terminal commands with table formatting, colored output, and JSON export.

bash
# Install globally (or use npx cryptyx)
npm install -g @cryptyx/mcp-server

# Save your API key
cryptyx config set-key cx_your_api_key

# Full state snapshot
cryptyx snapshot

# Factor scores for an asset
cryptyx factors BTC

# Top signals by information coefficient
cryptyx signals top

# Backtest a signal
cryptyx signals backtest TR_MOMO_CONT_14D --from 2025-04-20 --to 2026-04-20

# Raw JSON output (pipe to jq, scripts, cron)
cryptyx market --json

Run cryptyx --help for the full command list. Every command supports --json, --api-key, and --api-url flags.

Programmatic (TypeScript)

typescript
import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";

const transport = new StdioClientTransport({
  command: "npx",
  args: ["@cryptyx/mcp-server"],
  env: { CRYPTYX_API_KEY: process.env.CRYPTYX_API_KEY! },
});

const client = new Client({ name: "my-agent", version: "1.0.0" });
await client.connect(transport);

// List available tools
const { tools } = await client.listTools();
console.log(tools.map(t => t.name));

// Call a tool
const result = await client.callTool({
  name: "get_market_snapshot",
  arguments: { assets: "BTC,ETH,SOL" },
});
console.log(result.content);

Architecture

text
┌──────────────┐     stdio      ┌────────────────────┐    HTTPS     ┌──────────────┐
│              │ ◄──────────► │                    │ ──────────► │              │
│  AI Agent    │               │  @cryptyx/         │              │  CRYPTYX     │
│  (Claude,    │  MCP Protocol │  mcp-server        │  REST API    │  REST API    │
│   LangChain, │               │                    │              │              │
│   custom)    │               │  MCP tools         │              │  Neon        │
│              │               │  stdio transport   │              │  Postgres    │
└──────────────┘               └────────────────────┘              └──────────────┘

Tool Catalog (11 tools)

Asset Reads
get_coin_studyFree tier

Tactical (7–14 day) and structural (60–180 day) read for one asset: each reading's z-score against its own history, and forward returns after similar readings at 7, 30 and 90 days vs the base rate.

params: asset (required), asof_day?

get_episode_studyFree tier

Every prior instance of a move (breakdown, breakout, flush, capitulation) for one asset, and forward returns after each at 7, 30 and 90 days, with the t-stat and a random-entry test. Free accounts: 3 a day.

params: asset (required), preset or conditions, horizons?

check_coverageFree tier

Check which assets CRYPTYX covers.

params: assets (required)

get_data_livenessFree tier

When the data last updated, and when the next update is expected.

Market Scan
scan_metric_universe

Scan one metric across 180+ assets for z-score extremes on the latest published day.

params: metric_id (required), operator (required), threshold (required)

Strategy Research
get_featured_metrics

Metric directory: the metrics with the strongest historical relationship to later returns, market-wide or for one asset. A lead to test, not a result.

params: asset?

analyze_metric

Single-metric backtest: every time a reading crossed your threshold, and forward returns after each at 8 horizons vs the base rate.

params: metric_id (required), asset (required), operator (required), threshold (required)

analyze_metrics_composite

Multi-factor backtest: 2–4 conditions holding at once, every horizon vs the base rate.

params: asset (required), conditions (required)

Prices
get_price_history

Daily OHLCV candles for one asset.

params: asset (required), days?

get_live_prices

Latest spot prices, refreshed every 15 minutes. Pass the assets you need.

params: assets?

Also on the REST API and x402 pay-per-call, not on the connector in this release: quick scan, notable read, regime, setups and signals, composite scores, divergences, triggers and liquidity. See the API reference.

Factor Classes

CodeNameDescription
CORRCorrelationCross-asset correlation dynamics
EFFEfficiencyMarket efficiency and mean reversion
FLOWFlowCapital flow and fund movement
FUTFuturesDerivatives positioning and sentiment
OBOrder BookDepth, imbalance, and microstructure
OPTOptionsImplied vol, skew, and term structure
TRTrendPrice momentum and trend strength
VOLVolatilityRealized and implied vol dynamics