Prop Firm Rules MCP Server

prop-firm-rules: prop firm drawdown rules dataset + calculator (JS & Python)

Free, sourced prop-firm drawdown rules for traders and AI agents: dataset, JS/Python calculator, REST API and a no-auth MCP server.

MCP (Claude, Cursor, any MCP client), no key needed:

{
  "mcpServers": {
    "prop-firm-rules": {
      "type": "http",
      "url": "https://floorguard-kappa.vercel.app/api/mcp"
    }
  }
}

Claude Code: claude mcp add --transport http prop-firm-rules https://floorguard-kappa.vercel.app/api/mcp

Prefer to run it yourself? The repo ships a zero-dependency stdio server (mcp/server.mjs, Node 18+, works offline on the bundled dataset): {"mcpServers": {"prop-firm-rules": {"command": "npx", "args": ["-y", "github:kburrus64-max/prop-firm-rules"]}}}

▶ Live calculator (free, no sign-up, runs in your browser): floorguard-kappa.vercel.app/tools/drawdown-calculator.html

An open, sourced dataset of prop-firm risk rules (daily loss limit, max drawdown, static vs trailing, lock level, daily reset time) for FTMO, FundedNext, The5ers, E8 Markets, FundingPips, Alpha Capital, Blue Guardian, Topstep and Apex Trader Funding, plus a tiny zero-dependency drawdown calculator library in JavaScript and Python.

Every rule links to the firm's own help or rules page and carries a lastVerified date. Values that could not be confirmed on an official page are flagged verified: false or listed in unverifiedFields.

Try it without code: the free prop firm drawdown calculator on FloorGuard uses this same dataset and math in your browser.

Disclaimer: rules change. Prop firms update their rules often, and many let you pick different percentages at checkout. This data was checked on the date shown for each program. Always verify with your firm's current rules and your account dashboard before relying on any number here. Not affiliated with any prop firm; firm names are trademarks of their owners. Not financial advice.

Rules table (last verified 2026-10-04)

"Day-start" means the value at the firm's daily reset. "Higher of day-start balance/equity" means open profit at the reset raises the reference. Every limit is checked against equity (open P/L counts) unless the notes say otherwise. Per-program notes (for example Topstep's optional daily loss limit, Apex lock behaviour, Blue Guardian's Guardian Shield) are in data/firms.json.

The dataset also includes each firm's published copy-trading, multi-account and automation (EA/bot) policy (policy on each firm), with quotes and sources. unknown means the firm's pages did not answer the question.

What's in the repo

Usage

JavaScript (Node 18+)

git clone https://github.com/kburrus64-max/prop-firm-rules && cd prop-firm-rules
node --test js/test/

import { check, getFirm } from './js/index.mjs';

// FTMO 2-Step, $100k: balance closed yesterday at $102,000, equity now $101,200
const r = check('ftmo_2step', {
  size: 100000, currentBalance: 102000, currentEquity: 101200, todayPnl: -800,
  stopDistance: 20, valuePerUnit: 10, bufferPct: 10,   // 20-pip stop, $10/pip/lot, keep 10% buffer
});
console.log(r.daily.floor, r.daily.room);       // 97000 4200
console.log(r.drawdown.floor);                  // 90000 (static)
console.log(r.maxLots, r.bindingLimit, r.status); // 18.5 'daily' 'green'

console.log(getFirm('topstep').policy.remoteServerTrading); // 'banned'

In the browser, import js/calc.mjs and fetch data/firms.json:

import { evaluate } from './calc.mjs';
const data = await (await fetch('/data/firms.json')).json();
const program = data.firms.find(f => f.id === 'topstep').programs[0];
evaluate({ program, size: 50000, currentBalance: 50000, currentEquity: 50000, highWaterMark: 50500 }).drawdown.floor; // 48500

Python 3.9+

import sys; sys.path.insert(0, "python")
import prop_firm_rules as pfr

r = pfr.check("apex_intraday_eval", size=50000, current_balance=50000, current_equity=50900, today_pnl=900)
print(r["drawdown"]["floor"])   # 48900.0: the threshold trailed the open-profit peak

python -m unittest discover -s python/tests

evaluate() inputs and outputs

Output: daily and drawdown (floor, room, amount, status), maxLots, bindingLimit and an overall status of green (more than 50% of each allowance left), amber (50% or less) or red (20% or less, or breached).

Python uses the same names in snake_case (current_balance, high_water_mark, ...).

Use from AI agents

The same dataset is served live, free and read-only (no key, CORS open, ~60 requests/minute per IP) by FloorGuard:

• MCP server (Streamable HTTP, no auth): https://floorguard-kappa.vercel.app/api/mcp
  • Tools: list_firms, get_rules(program), check_drawdown(program, accountSize, currentEquity, ...)
  • Listed in the official MCP Registry as io.github.kburrus64-max/prop-firm-rules (see server.json).
  • Client config example: {"mcpServers": {"prop-firm-rules": {"type": "http", "url": "https://floorguard-kappa.vercel.app/api/mcp"}}}
• Local MCP server (stdio, no network needed): npx -y github:kburrus64-max/prop-firm-rules, or after cloning, node mcp/server.mjs. Same three tools and the same math as the hosted endpoint.
• REST API: /api/v1/firms, /api/v1/rules?program=ftmo_2step, /api/v1/check; OpenAPI 3.1 at /api/v1/openapi.json.
• A2A agent card: /.well-known/agent-card.json
• llms.txt: /llms.txt

Informational only, not financial advice; verify every value with the firm.

Static vs trailing drawdown in 30 seconds

• Static: the floor is fixed at initial balance minus the allowance (FTMO 2-Step: $90,000 on $100k).
• Trailing end-of-day: the floor follows your highest end-of-day balance (Topstep MLL, FTMO 1-Step, Apex EOD).
• Trailing closed balance: follows your highest closed balance (E8 dynamic drawdown).
• Trailing intraday: follows your peak including open profit, in real time (Apex intraday).
• Lock: some trailing floors stop rising at a level such as the starting balance (Topstep, E8).

Longer explanations with sources: trailing vs static drawdown, FTMO daily loss rule, same 5% daily loss, different math.

Corrections welcome

Found a rule that changed? Open an issue with the firm's official URL and the quoted text (see CONTRIBUTING.md). Please don't submit values without an official source.

Limitations

• Commissions, swaps, slippage and server-side rounding are not modelled; firms' platforms can differ by a few dollars.
• Account-size lists marked sizesVerified: false were not confirmed on an official page.
• Programs with features this model does not cover (e.g. Blue Guardian funded-account Guardian Shield, consistency rules, news restrictions, payout rules) are noted in notes but not calculated.

License

MIT. The dataset is a compilation of facts from the firms' public pages, with links back to them. Built by the team behind FloorGuard, a prop-account risk guard in development.