AI Connection
Connect Claude to your data
Add AnalyticsFX as a connector in Claude and ask about your own trades, backtests, forward-testing accounts, seasonality, COT and fundamentals in plain language. No code.
Connect in one click
- 1. In Claude, open
Settings → Connectors → Add custom connector. - 2. Paste this URL:
https://www.analytics-fx.com/api/mcp- 3. Click
Connect. A browser window opens, you sign in to AnalyticsFX and approve. That is it.
After connecting, just ask Claude things like "summarize my last 50 trades" or "where am I losing money?"
What you can ask
- Summarize my recent trades and win rate.
- Where am I losing money, by session or symbol?
- How are my forward-testing positions doing right now?
- What is the seasonal pattern for EURUSD this month?
- What does COT positioning look like for the euro?
Your data is visible only to you. Most tools only read; the ones that write are marked, and they only ever touch your own journal, backtests and forward-testing accounts, including placing and closing trades on a demo account.
Tools
Every tool is scoped to the connected account. Read tools never expose another user's data; write tools change only your own journal, backtests and forward-testing accounts, and are clearly marked.
get_seasonalityreadCompiled seasonal statistics for an instrument symbol: monthly average/median return, win rate, average annual return, and best/worst/most-consistent months. Same data as the /api/v1/seasonality API.
get_cotreadCommitments of Traders positioning for a currency or instrument code. Same data as the /api/v1/cot API (history capped at ~3 months).
get_crypto_sentimentreadThe crypto Fear and Greed index (alternative.me), Bitcoin and Ethereum only, daily since 2018-02-01: the 0-100 level with its classification, the change over a week and a month, and a short reading. A sentiment and attention proxy, not an on-chain or valuation measure. The index is market wide, so BTC and ETH report the same value on a date.
get_fundamentalsreadFundamentals snapshot for an equity symbol.
list_supported_symbolsreadList the symbols and currencies supported per dataset (seasonality, cot, stocks).
get_gexreadGamma exposure (GEX) profile for an options-listed symbol: per-strike call/put/total GEX and key levels (gamma flip, max gamma). A symbol the GEX service does not cover answers invalid_request with its reason, and is not charged.
get_zetareadLatest Zeta regime/risk model for an asset (SPY, QQQ, DAX): regime, drift, risk and series.
get_mlreadML predictions for a currency: ATR (volatility) forecast and macro model output. USD maps to DXY.
get_newsreadEconomic calendar events, newest first. Optional currency and ISO date range.
get_calendarreadThe economic calendar since 2007 for the eight major currencies: every release with its ACTUAL, its FORECAST and the surprise between them, plus what is still to come.
get_policy_ratesreadCentral bank policy rates (Fed, ECB, BoE, BoJ, SNB, RBA, RBNZ, BoC): the rate in force on a date with when it last changed, recent changes and the next scheduled decision.
get_macro_seriesreadStored macro series from official keyless sources, for research and scripts: government bond yields (US Treasuries, German Bunds, UK gilts, JGBs, Government of Canada and Swiss Confederation bonds), US real yields and breakevens, the VIX, SOFR and €STR, the Fed's broad dollar index (not ICE's DXY), WTI and Brent spot, CPI, unemployment and business confidence for USD, EUR, GBP, JPY, CHF, CAD, AUD and NZD, and liquidity (Fed and ECB balance sheets, the Treasury General Account, overnight reverse repo and US net liquidity). For each series: the latest value on or before a date, its changes over 1, 5 and 20 observations, its unit and frequency, latestReadableFrom (when the latest value was public, which is when a backtest first reads it; US monthly figures carry their ALFRED first-release date), the first stored date, the source URL and its attribution line, which you quote with the numbers (FRED, Cboe and the New York Fed require it).
get_screenerreadRanked seasonal windows (the seasonality screener candidates) for a symbol.
run_montecarloreadRun a seeded Monte Carlo simulation of a trading strategy and return summary stats (win/loss %, mean max drawdown, prop-firm phase pass rates).
list_journal_accountsreadList the connected user's own journal accounts (mt and mt5).
get_tradesreadClosed trades for one of the user's mt5 accounts. profit is already net of commission and swap.
get_open_positionsreadThe user's open positions and pending orders across their MT5 accounts, with floating P/L REPRICED against the live feed rather than left at the last bridge sync, plus each account's balance, equity and free margin. The rows themselves (SL/TP, ticket, open time) are from the last sync; the price and the P/L are current. profit is floating and excludes swap and commission.
analyze_account_performancereadSummary of one of the user's mt5 accounts: trade count, win rate, net P/L and where they lose money.
get_backtestsreadList the user's own backtests in a clean public format, or fetch one with its trades by passing id. Same data as the /api/v1/backtests API. Internal engine fields are never exposed.
add_backtest_tradeswriteAppend trades to one of the user's existing backtests (by id from get_backtests or create_backtest_session). Owner-scoped. Trade times are ISO-8601 or epoch ms; unknown fields are ignored. Returns the updated backtest with its full trade list.
remove_backtestwriteDelete one of the user's backtests by id; its trades are removed too. Owner-scoped.
create_backtest_sessionwriteCreate a new backtest session (symbol + timeframe replay with its own starting capital). contractSize defaults from the instrument's broker contract size when omitted (never assume the FX lot of 100000; that breaks index sessions).
get_backtest_sessionreadOne of the user's backtest sessions with its key stats (trades, net P&L, win rate, profit factor). Use get_backtest_trades for the trade list.
list_backtest_sessionsreadThe user's backtest sessions (summary, newest first), paginated.
get_backtest_tradesreadClosed trades for one of the user's backtest sessions.
delete_backtest_sessionwriteDelete one of the user's backtest sessions by id; its trades are removed too.
get_backtest_analyticsreadFull analytics overview for one of the user's backtest sessions: expectancy, win rate, profit factor, drawdown, average R/R and discipline flags (revenge trading, size creep, overtrading). Computed by the same engine as the app's Overview tab.
get_forward_accountsreadList the user's forward-testing accounts (demo accounts traded against live data), or fetch one with its full trade list by passing id. Stats cover the closed trades (net P&L, win rate, profit factor, balance) and the live state: floating P&L, equity, used margin, and how many trades are open or pending. commission is what the account charges per close (perLot, perOrder, percent, a bare number, or null for none), already taken off each closed trade's pnl. Same data as the /api/v1/forward/accounts API.
get_forward_tradesreadThe trades on one of the user's forward-testing accounts, newest first, each with the current floating P&L, live price and used margin of the position. One call answers 'how are my positions doing'; no per-trade lookup is needed for that. Owner-scoped.
get_forward_tradereadEverything about one forward trade: its live valuation, the scale-out ladder, the break-even trigger, the notes and confluences, and the audit trail of what the server did to it and why. get_forward_trades already carries the floating P&L of every position, so this is for the detail and the audit trail.
open_forward_tradewritePlace an order on one of the user's active forward-testing accounts. This moves a real position on that demo account. A market order fills at the last completed 1-minute close, the same series the server settles against, so no price is sent for one; a limit order rests at the given price until the server fills it. Stops, targets and stop-outs are then handled server-side every minute.
modify_forward_tradewriteChange the risk levels or annotations on one of the user's forward trades. Only stopLoss, takeProfit, takeProfitLadder, breakEvenTrigger, notes and confluences can be changed; size, entry, exit, P&L and status are decided server-side and are never editable.
close_forward_tradewriteClose one of the user's open forward trades at market, or cancel it when it is still an unfilled order. The exit price comes from the last completed 1-minute close and the realized P&L is recomputed server-side, so no price is accepted.
connect_mt5_accountwriteConnect an MT5 account to the user's journal so its trades sync. Credentials are only used to start the sync and are never returned.
add_trade_screenshotwriteAttach a screenshot to a specific trade in the user's journal. Provide the image as base64.
add_journal_tradewriteAdd a manual trade to one of your custom journal accounts (only custom accounts accept manual trades).
remove_journal_tradewriteRemove a manual trade (by its id) from one of your custom journal accounts.
Security and scope
- Auth uses OAuth 2.1 with PKCE; your client discovers the endpoints from
/.well-known/oauth-authorization-server. An API key from /dashboard/api also works as a bearer token. - Journal, backtest and forward-testing tools are owner-scoped: an assistant can only read or modify your own accounts and trades, never anyone else's.
- Read tools are side-effect free. Write tools are marked and change only your own data: your journal (connect_mt5_account, add_trade_screenshot, add_journal_trade, remove_journal_trade), your backtests (add_backtest_trades, remove_backtest), and your forward-testing accounts (open_forward_trade, modify_forward_trade, close_forward_trade), which place and close trades on a demo account.
- Usage is metered the same way as the REST API.
