TradingCalc MCP Server
An LLM guesses trading maths. This server computes it: the same inputs always return the same numbers, every result carries a signature you can check, and the core formulas are tested against public vectors (live count in the badge above).
Your question MCP tool call Result
+------------------+ +---------------------------+ +---------------------------+
| "Risk $200 on | ► | workflow.run_forex_ | ► | 0.67 lots, plus an ECDSA |
| EUR/USD with a | | position_size_live | | P-256 signature you can |
| 30 pip stop" | | (deterministic maths) | | verify offline |
+------------------+ +---------------------------+ +---------------------------+
Ask Claude, Cursor or any MCP client trade questions and get exact numbers back, not guesses:
"What should a BTC call at a $90k strike, 30 days out, with 55% implied volatility be worth?" "How many lots of EUR/USD should I trade to risk $200 with a 30-pip stop?" "What's my 95% Value at Risk on this monthly return series?" "What's my PnL if I buy 0.5 BTC at $80k and sell at $95k with 5x leverage?"
What it covers (75 tools)
Access via MCP (Claude Desktop, Cursor, VS Code and other clients) or a plain HTTP POST to the MCP endpoint. No signup. The server is remote, so nothing to install beyond a one-line client config; only the Claude Desktop bridge below needs Node.js 18 or newer.
Endpoints
MCP transport: Streamable HTTP (MCP spec 2024-11-05)
Quick Start
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"tradingcalc": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://tradingcalc.io/api/mcp"]
}
}
}
Cursor / VS Code
{
"tradingcalc": {
"url": "https://tradingcalc.io/api/mcp"
}
}
Agent Skill
For coding agents that support the skills.sh ecosystem (Claude Code, Cursor,
GitHub Copilot, and others) - installs a SKILL.md that teaches the agent when to reach for these
tools instead of estimating trade math itself:
npx skills add SKalinin909/tradingcalc-mcp
Direct HTTP
curl -X POST https://tradingcalc.io/api/mcp \
-H "Content-Type: application/json" \
-d '{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "workflow.run_liquidation_safety",
"arguments": {
"side": "long",
"entryPrice": 95000,
"leverage": 10
}
}
}'
Example prompts
After connecting, just ask naturally: the AI picks the right tool automatically:
Options payoff
"What does my BTC call payoff at $90,000? Strike $85,000, premium 0.02 BTC, long 1 contract."
Black-Scholes / IV check
"Is Deribit's BTC-27FEB27-90000-C fairly priced right now?"
Covered call yield
"I hold BTC bought at $80,000. What annualized yield do I get selling a $85,000-strike call for 0.02 BTC, 30 days out?"
Forex pip value
"What's a pip worth on 2 lots of EUR/USD, in my USD account?"
Forex position size
"How many lots should I trade to risk $200 on EUR/USD with a 30-pip stop?"
Forex margin level
"My equity is $5,000 and used margin is $1,800. How close am I to a margin call?"
Value at Risk
"What's my 95% Value at Risk and expected shortfall on this monthly return series? [paste returns]"
Sharpe ratio
"What's the Sharpe ratio of these monthly returns, corrected for autocorrelation?"
Cointegration
"Are these two price series cointegrated? Is there a pairs trade?"
Odds converter
"What odds does a 35% probability work out to?"
Market-implied odds
"What does the market think BTC will be worth by year end?"
Prediction market edge
"I think this event is 60% likely but the market prices it at 40%. Should I bet, and how much with a $10k bankroll?"
Token safety check
"Is this Solana token a rug pull risk? Mint: DezXAZ8z7PnrnRJjz3wXBoRgixCa6xjnB7YaB1pPB263"
Swap price impact
"How much slippage will I eat swapping 50,000,000 BONK to USDC?"
Bonding curve
"How many tokens do I get buying with 1 SOL on a pump.fun curve that's already raised 20 SOL?"
Market cap comparison
"If I put $1,000 into BONK and it reaches JUP's market cap, what's it worth?"
Wallet flag check
"Is this wallet address flagged for anything? [address]" (Solana or EVM)
Trade P&L
"I bought 0.5 BTC at $80,000 and want to sell at $95,000 with 5x leverage. What's my net profit after fees?"
Position sizing
"I have a $10,000 account and want to risk 1% going long BTC at $83,000 with a stop at $81,000. How many coins should I buy?"
Liquidation check
"Long ETH at $3,200 with 10x leverage, where do I get liquidated?"
Full pre-trade check
"Analyze this setup: long BTC at $83,000, stop $81,000, target $90,000, $10k account, 1% risk, 5x leverage. Is it worth taking?"
Funding cost
"I'm holding 0.5 BTC long on Bybit at $83,000 with 0.01% funding rate. How much will funding cost me over 3 days?"
Carry trade
"Is this carry trade worth it? Long on Bybit at 0.01% funding, short on Binance at 0.05%, $50k notional, 30 days."
DCA average entry
"I bought BTC at $78k (0.2 BTC), $80k (0.3 BTC), and $82k (0.1 BTC). What's my average entry and breakeven?"
Scale-out plan
"I'm long 1 BTC from $80k. I want to close 30% at $88k, 40% at $92k, 30% at $96k. What's my total P&L?"
Tools (75)
Tool naming follows the workflow.run_* / primitive.* / system.* namespace convention.
Old flat names (pnl, liquidation, etc.) are accepted for backward compatibility. All tools are
free via MCP, no signup; 100 calls/day anonymously, 200/day with a free API key.
Full tool-by-tool reference (every input/output schema, request/response examples, per-tool descriptions): docs.tradingcalc.io/api
Formulas normalized across 17 exchanges: Binance, OKX, Bybit, Aster, Hyperliquid, MEXC, KuCoin, Gate, Deribit, Kraken, HTX, WOO, Phemex, Blofin, Backpack, dYdX, Polymarket Perps.
Rate Limits
The Service is free. Need a higher-limit key? Email [email protected].
Pass key as: Authorization: Bearer <your-api-key>
Self-Verification
Agents can verify all the canonical test vectors before trusting results:
{
"jsonrpc": "2.0", "id": 1,
"method": "tools/call",
"params": { "name": "system.verify", "arguments": {} }
}
Response: { "status": "pass", "passed": 43, "failed": 0, "total": 43 }
Live proof: tradingcalc.io/verify
Signed Responses
system.verify proves the formulas are correct. It doesn't prove the specific response you got
wasn't altered by a proxy, cache, or MITM in between. Every tools/call result carries a second
content block signed with ECDSA P-256, plus X-TradingCalc-Signature/Kid/Signed-At headers.
Call system.pubkey (or GET /api/mcp/pubkey) for the public key (PEM + JWK) and the canonical
string format needed to verify offline, no callback required.
Once a day the verification state is also signed and its SHA-256 is written to Solana mainnet in a Memo transaction (tcalc1 {date} {hash}), each record carrying the hash of the day before. Read the records at GET https://tradingcalc.io/api/anchors; the latest record and how to verify it yourself are at tradingcalc.io/verification.
Use Cases
- Options and forex desks: price an option, size a forex position, or check margin level in one exact call
- Quant research: Sharpe, VaR, cointegration and drawdown statistics from a return series, cross-checked against reference libraries
- Trading bots: check liquidation price before every trade
- AI agents: deterministic risk calculations without hallucination risk
- Multi-agent systems: drop-in risk management agent in analyst + risk + execution pipelines
- Dashboards: embed calculations programmatically
Why deterministic?
LLMs asked directly give plausible but potentially wrong numbers. TradingCalc MCP returns exact calculations: same inputs always produce the same outputs. No hallucination risk for financial data.
Risk Agent Wrapper
examples/risk-agent-wrapper.ts: a drop-in TypeScript wrapper for risk-gated trade execution.
Integrates with any agent framework (ElizaOS, CrewAI, AutoGen, Hummingbot, Freqtrade).
import { RiskAgent, preTradeGate } from './examples/risk-agent-wrapper';
const agent = new RiskAgent({ apiKey: 'tc_your_key', minLiqDistancePct: 3.0 });
const result = await agent.evaluate({
symbol: 'BTCUSDT', exchange: 'bybit',
side: 'long', entry_price: 83000, stop_loss: 81000,
account_balance: 10000, risk_pct: 1, leverage: 5,
funding_rate: 0.0001, hold_hours: 24,
});
if (result.approved) {
// execute trade: result.recommended_size, result.liquidation_price
} else {
console.log('Rejected:', result.rejection_reason);
}
// Binary gate for execution bots
const ok = await agent.isSafe({ symbol: 'ETHUSDT', side: 'short', ... });
// Standalone function (minimal integration)
const { approved, size, liqPrice } = await preTradeGate({ ... }, 'tc_your_key');
TypeScript SDK
For code-first integrations, use tradingcalc-sdk instead of raw JSON-RPC:
npm install tradingcalc-sdk
import { TradingCalcClient } from 'tradingcalc-sdk';
const tc = new TradingCalcClient({ apiKey: 'tc_your_key' });
// Workflows: orchestrated decisions
const check = await tc.workflows.preTradeCheck({ side: 'long', entry_price: 83000, leverage: 5, funding_rate: 0.0001, account_balance: 5000 });
// Primitives: single formula
const avg = await tc.primitives.averageEntry({ symbol: 'BTCUSDT', input: { fills: [{ price: 83000, quantity: 0.1 }] } });
// System
const report = await tc.system.verify();
tc.call() is available for raw MCP access. Full docs: npmjs.com/package/tradingcalc-sdk
Links
- For agents: tradingcalc.io/for-agents
- Full docs (API & MCP reference, methodology, architecture, changelog): docs.tradingcalc.io
- Verification proof: tradingcalc.io/verify
- Web calculators: tradingcalc.io