MCP · STREAMABLE HTTP · 77 TOOLS

Deterministic trading math for AI agents

TradingCalc is the computation layer your agent calls for exact options, forex, risk and statistics, prediction-market, on-chain, and futures math. Every result is signed, and every formula is published. Same inputs always produce the same outputs. Every formula is traceable to exchange documentation and regression-tested.

https://tradingcalc.io/api/mcp

77 tools·11 suites·17 exchanges·self-verifiable
tools/call · workflow.run_pnl_planning
$ 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_pnl_planning",
"arguments":{"side":"long",
"entryPrice":83000,"exitPrice":85000,
"size":0.1,"feeOpenPct":0.0002,
"feeClosePct":0.0005}}}'
 
// result.content[0].text
{
"pnl": 194.09,
"pnlPct": 0.02339,
"fees": 5.91,
"gross": 200.00
}
01WHY DETERMINISTIC

Why deterministic matters

LLM ASKED DIRECTLY
Right 97% of the time in our October 2026 test. The misses look like the hits, and the answer can change between calls.
TRADINGCALC MCP CALL
Exact number from a published, tested formula, in a signed result. Same inputs, same output.

The model does the reasoning. TradingCalc does the math.

02TOOL SUITES

77 MCP tools across 11 suites

Options, Forex, Risk/Stats, Prediction Markets, On-Chain, Crypto Spot, and for perpetual futures Trade Planning, Risk & Margin, Funding & Carry, Integrated Decision and Market Structure: every tool is callable by name over the same JSON-RPC endpoint.

FULL TOOL CATALOG →
03BUILT FOR

Built for

AI agents & assistants
Call exact trading math instead of improvising calculations in natural language.
Trading bots
Verify risk, sizing, and target structure with deterministic calls before every order.
Risk management systems
Embed into internal risk rails where same inputs must always produce the same outputs.
Multi-agent pipelines
Use as the deterministic layer inside analyst → risk → execution workflows.
Dashboards & analytics
Surface exact outputs inside operator dashboards and decision-support interfaces.
Algo trading strategies
Reusable computation substrate across planning, carry, hedge, and decision safety.
04AGENT WORKFLOWS

Example agent workflows

Pre-trade risk check
  1. 1.Call `workflow.run_risk_reward` → R:R ratio, position size, liquidation, P&L at stop and target, verdict
  2. 2.Call `workflow.run_funding_cost` → cumulative funding cost over hold period
  3. 3.Call `workflow.run_pre_trade_check` → full decision: sizing + liquidation + breakeven + funding in one call
DCA and scale-out
  1. 1.Call `workflow.run_dca_entry` with fill levels → avg entry, breakeven, per-level contribution
  2. 2.Call `workflow.run_liquidation_safety` → liquidation distance from averaged entry
  3. 3.Call `workflow.run_scale_out` with exit levels → P&L per exit, weighted avg exit, overall ROI
Funding carry analysis
  1. 1.Call `workflow.run_carry_trade` → net yield, gross/net profit, breakeven days, verdict
  2. 2.Call `workflow.run_funding_breakeven` → price move needed to cover funding + fees
  3. 3.Call `workflow.run_funding_arbitrage` → annualized arb yield between two exchanges
Scenario planning
  1. 1.Call `workflow.run_scenario_planning` with price targets → full P&L table
  2. 2.Call `workflow.run_exit_target` → exact price to hit target ROE
  3. 3.Call `primitive.hedge_ratio` → perp short size to hedge the position
05SELF-VERIFICATION

Self-verification

Three independent proofs, not one claim: system.verify confirms the formulas are correct (46/46 test vectors, callable via MCP before you trust a result), and every tools/call response is signed with ECDSA P-256 so you can confirm the specific response you got wasn't altered in transit, not just that the formula behind it is right. The third proof is public: once a day the verification state is signed and its hash is written to Solana, so a past day cannot be quietly rewritten.

06ACCESS METHODS

Access methods

CLAUDE DESKTOP
claude_desktop_config.json
{
"mcpServers": {
"tradingcalc": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://tradingcalc.io/api/mcp"]
}
}
}

Free tier: no signup required for anonymous use. Cursor / VS Code config, direct HTTP calls, and API-key authentication for higher limits are in the full API & MCP reference.

07FAQ

Frequently asked questions

Q.01
Is TradingCalc an AI?
No. It’s a deterministic computation layer. Same inputs always produce the same outputs.
Q.02
What does “deterministic” mean?
No randomness. No estimation. Same inputs, same outputs, every time.
Q.03
Does it work in Claude Desktop / Cursor / VS Code?
Yes. Quickstarts for all three are above.
Q.04
Can I use it without an API key?
Yes. 100 calls/day anonymously, no signup required.
Q.05
How do I verify results are correct?
Call system.verify via MCP, or visit /verify for the live proof page.
Q.06
How do I verify a response actually came from TradingCalc?
Every tools/call result includes a second content block signed with ECDSA P-256. Call system.pubkey for the public key and canonical string format to verify offline.
Q.07
Which exchanges are supported?
17 exchanges, including Binance, Bybit, OKX, Hyperliquid, Kraken, Deribit, and dYdX. Live prices and funding rates come from each exchange, and each exchange page lists that exchange's base trading fees. The formulas themselves are general, so an exchange's own liquidation engine (size tiers, liquidation fee) can show a slightly different figure.
Q.08
Can I use it in an automated trading bot?
Yes. That’s a primary use case. MCP and GPT Actions both work.
Q.09
What is an integrated decision tool?
A single call that orchestrates multiple primitives. pre_trade_check runs position sizing, liquidation, breakeven, and funding cost together and returns a structured safety classification.
Q.10
Is it free?
Yes. 100 calls/day anonymously; a free API key raises it to 200/day. No paid plan.

Start in 30 seconds

No signup required. Add to Claude Desktop or Cursor using the config above, or call the endpoint directly.