FxSocket Python SDK

Typed Python client for the FxSocket API. Connect your MetaTrader 4 / 5 accounts, then place trades, read market data, and stream live updates over REST and WebSocket — with mirrored synchronous and async interfaces.

Features

• Account management — link, list, fetch, and disconnect MT4/MT5 accounts.
• Private servers — list your dedicated hosting servers and manage the accounts on them.
• Read-only keys — mint, scope, rotate and revoke fxs_ro_… keys for dashboards and monitors.
• Trading — market & pending orders, modify, close, close-all, plus margin/profit calculators.
• Multi-account trading — one order, or one close, fanned out to several accounts in a single request, with idempotency keys for safe retries.
• Wallet — read your prepaid balance, pending top-ups and upcoming charges.
• Market data — quotes, symbol specifications (incl. commission rules & trading sessions), OHLC history, account state & info.
• Live streaming — ticks, bars, account, positions, trades, and terminal status over WebSocket, with automatic reconnect + subscription replay.
• Sync and async — Client / AsyncClient, method-for-method mirrors.
• Typed — Pydantic v2 models throughout; ships py.typed.
• One interface for MT4 and MT5 — platform differences handled for you.

Install

pip install fxsocket

Requires Python 3.10+.

Quickstart

from fxsocket import Client

with Client(api_key="fxs_live_…") as fx:        # or set FXSOCKET_API_KEY
    account = fx.accounts.list()[0]
    term = fx.terminal(account)

    print("equity:", term.account_summary().equity)
    print("EURUSD:", term.quote("EURUSD").ask)

Authentication

Every call uses your FxSocket API key (fxs_live_…), from the dashboard. Pass it explicitly, or set the FXSOCKET_API_KEY environment variable and call Client() with no arguments.

from fxsocket import Client

with Client(api_key="fxs_live_…") as fx:
    for account in fx.accounts.list():
        print(account.platform, account.nickname, account.status)

Managing accounts

from fxsocket import Client

with Client(api_key="fxs_live_…") as fx:
    # Link a new account (platform defaults to MT5).
    account = fx.accounts.create(
        platform="mt5", server="ICMarkets-Demo", login=1150125, password="…",
    )

    # Poll until it's connected.
    account = fx.accounts.get(account.id)
    print(account.status)                  # connecting → connected

    # Where this account's terminal API lives (empty until provisioned).
    print(account.rest_url, account.ws_url)

    # Move the trade expert to a specific chart symbol (the terminal
    # restarts on it — poll until connected again). "" reverts to automatic.
    account = fx.accounts.update(account, trade_ea_symbol="EURUSDm")

    fx.accounts.delete(account.id)         # unlink

An account's terminal can be routed through an outbound proxy at link time. The proxy is verified first — an unreachable one raises ConnectFailedError with code == "proxy_unreachable" and nothing is created. The address, type and local port are readable on the Account; the credentials never come back.

account = fx.accounts.create(
    server="ICMarkets-Demo", login=1150125, password="…",
    trade_ea_symbol="EURUSDm",                 # optional, broker-exact
    proxy_address="10.0.0.5:1080", proxy_type="socks5",
    proxy_auth="user:secret", proxy_local_port=1080,
)

Everything is also available on AsyncClient:

from fxsocket import AsyncClient

async with AsyncClient(api_key="fxs_live_…") as fx:
    accounts = await fx.accounts.list()

Read-only keys

A read-only key (fxs_ro_…) can call every GET endpoint but nothing that mutates state. fx.readonly_keys manages named ones; each has a scope — all sees every account, selected only the accounts attached to it (everything else is absent from lists, 404 by id and 401 at the terminal). Passing accounts implies selected.

from fxsocket import Client, KeyScope

with Client(api_key="fxs_live_…") as fx:
    key = fx.readonly_keys.create(name="dashboard", accounts=[account])
    print(key.key)                          # plaintext, returned on every read

    key = fx.readonly_keys.update(key, name="ops-dashboard")
    key = fx.readonly_keys.rotate(key)      # new secret, same name & scope
    fx.readonly_keys.delete(key)            # revoke — immediate, irreversible

    for k in fx.readonly_keys.list():
        print(k.name, k.scope == KeyScope.ALL, k.last_used_at)

Terminals are started with the exact set of read-only keys they accept, so creating, re-scoping, rotating or revoking a key restarts the terminals of every account in its scope — each goes briefly offline, typically a few minutes. Renaming is free. Key management itself always needs the full fxs_live_… key, even for reads, because the replies contain plaintext key values.

Trading & market data

fx.terminal(account) returns a REST client bound to that account's terminal (resolved from account.rest_url, whether it's a shared pod or a private droplet):

from fxsocket import Client

with Client(api_key="fxs_live_…") as fx:
    account = fx.accounts.get("…")         # must be connected (has a terminal)
    term = fx.terminal(account)

    summary = term.account_summary()       # balance, equity, margin, …
    quote = term.quote("EURUSD")           # latest tick
    bars = term.price_history("EURUSD", "M5")   # recent OHLC bars

    result = term.order_send(              # market buy
        symbol="EURUSD", operation="Buy", volume=0.10,
        stop_loss=1.07, take_profit=1.10,
    )
    if result.success:
        term.order_modify(result.order, take_profit=1.12)   # None keeps the SL
        term.order_close(result.order)

Every order call returns an OrderResult. A 200 only means the terminal answered — check the body: success is true for retcode 10009 (done) or 10008 (placed), and outcome classifies the result as applied / no_change / partial / rejected (compare against OrderOutcome).

no_change (retcode 10025) is a benign, idempotent no-op — the requested SL/TP/price already match — so it's safe to treat as applied even though success is False. For idempotent SL/TP management (e.g. re-sending after a lost confirmation), send absolute values and gate on result.is_effective (true for both applied and no_change):

res = term.order_modify(ticket, stop_loss=1.0850)
if res.is_effective:        # applied now, or already in effect
    ...

There's also a panic button. close_all() closes every open position in one trade-EA pass — optionally filtered by symbol and/or magic (magic=0 matches manually-opened orders), and delete_pending=True also deletes matching pending orders. It returns a CloseAllSummary with per-ticket results. On a 504 the pass continues inside the terminal — check opened_orders() before acting again rather than re-sending:

summary = term.close_all(symbol="EURUSD", delete_pending=True)
if summary.failed:
    for r in summary.results:
        if not r.success:
            print(r.ticket, r.retcode, r.retcode_description)

Inputs are validated client-side before they're sent. One guard worth knowing: in order_modify, a literal stop_loss=0.0 would remove your stop-loss, so it's rejected — pass clear_stop_loss=True to remove one deliberately, while None (the default) keeps the current value.

MT4 and MT5 share one interface, timeframes included. Both accept the same nine: M1, M5, M15, M30, H1, H4, D1, W1, MN1 (or 1min, 5min, 15min, 30min, 1h, 4h, 1d, 1w, 1month). MT5's other periods, such as H2 or H12, aren't served by the terminal API, so anything else raises ValidationError before any request.

MT4 history note: on MT4, price_history with from_/to bounds (or the D1 timeframe) can fail server-side with CopyRates failed when the terminal hasn't loaded that history. Calling price_history(symbol, timeframe) without date bounds returns the most recent bars reliably.

Multi-account trading

fx.orders sends one order — or one close — to several accounts in a single request, through the management API rather than each terminal. Every leg inherits defaults and may override any of it, so "same trade, three accounts, three lot sizes" stays short. A leg can be an OrderLeg, a plain dict, or just the account (object or id) when defaults say everything else.

from fxsocket import Client, OrderDefaults, OrderLeg

with Client(api_key="fxs_live_…") as fx:
    accounts = [a for a in fx.accounts.list() if a.has_terminal]

    result = fx.orders.send(
        [OrderLeg(account_id=a, volume=0.1 * (i + 1)) for i, a in enumerate(accounts)],
        defaults=OrderDefaults(symbol="EURUSD", operation="buy", stop_loss=1.07),
        idempotency_key="signal-4711",
    )
    for account, leg in zip(accounts, result.results):
        print(account.nickname, leg.status, leg.order, leg.message)

Validation is all-or-nothing, execution is not. A malformed leg raises ValidationError before anything is sent (the SDK checks what the API checks — symbol / operation / volume present, a price for pending orders, …). Once dispatched the legs are independent: you get a BatchOrderResult with a positional results list — zip it with what you sent rather than matching on account_id, which repeats when several legs target one account. Each OrderLegResult.status is only trustworthy as filled; a timeout leg may have executed (is_unknown), so never blind-retry it. result.all_filled, filled_legs, failed_legs and unknown_legs roll this up.

Symbols are per broker (EURUSD, EURUSD.sd, EURUSDm), so a single defaults symbol across mixed brokers will partly fail by design — that is what a leg-level symbol is for.

Idempotency. Pass an idempotency_key (any opaque string, ≤ 128 chars) whenever a retry is possible: replaying the identical batch with the same key returns the stored reply (idempotent_replay=True) and sends nothing. Keys are remembered for 15 minutes. A key that is still in flight, was already used for a different body, or can't currently be guaranteed raises IdempotencyError — nothing is sent in any of those cases.

Closing works by selector, not by ticket: each account is matched against symbol (plus optional side, kind, magic), the backend resolves the tickets from that account's open orders and closes them. symbol is required — type "*" to mean every symbol; it is never implied. symbol_match="base" lets one selector reach EURUSD, EURUSD.sd and EURUSDm across brokers. kind defaults to position, so a routine close does not also delete resting pending orders.

from fxsocket import CloseDefaults, CloseLeg

closed = fx.orders.close(
    [
        CloseLeg(account_id=accounts[0], side="long", volume=0.05),   # partial
        CloseLeg(account_id=accounts[1], tickets=[123456, 123457]),  # explicit
        accounts[2],                                                 # defaults only
    ],
    defaults=CloseDefaults(symbol="EURUSD", symbol_match="base"),
    idempotency_key="flatten-4711",
)
for leg in closed.results:
    print(leg.account_id, leg.status, f"{leg.closed}/{leg.matched}")
    for ticket in leg.results:
        print("   ", ticket.ticket, ticket.status, ticket.retcode_description)

nothing_matched is a normal answer, not an error. Per-account status compares against CloseLegStatus and per-ticket against ClosedTicketStatus; skipped tickets were never sent, timeout ones may well have closed. An idempotency_key matters most for partial closes, where a blind retry genuinely over-closes.

Both calls need the full fxs_live_… key — a read-only key raises ForbiddenError.

Streaming (WebSocket)

Subscribe to live ticks, bars, account, positions, trades, and terminal status. Streaming is async-first; a synchronous wrapper is provided too. A dropped connection auto-reconnects and replays active subscriptions (auto_reconnect=True by default).

import asyncio
from fxsocket import AsyncClient, Tick, Bar, AccountUpdate

async def main():
    async with AsyncClient(api_key="fxs_live_…") as fx:
        account = await fx.accounts.get("…")
        async with fx.stream(account) as s:
            await s.subscribe_prices("EURUSD")
            await s.subscribe_bars("EURUSD", "M5")
            await s.subscribe_account()
            async for event in s:
                match event:
                    case Tick():
                        print(event.symbol, event.data.bid, event.data.ask)
                    case Bar():
                        print(event.symbol, event.timeframe, event.data.close)
                    case AccountUpdate():
                        print("equity", event.data.equity)

asyncio.run(main())

Synchronous equivalent:

from fxsocket import Client, Tick

with Client(api_key="fxs_live_…") as fx:
    with fx.stream(fx.accounts.get("…")) as s:
        s.subscribe_prices("EURUSD")
        for event in s:
            if isinstance(event, Tick):
                print(event.data.bid, event.data.ask)

Trade events

A TradeUpdate carries the full deal: commission, swap, magic and a real comment alongside profit (bridges MT5 0.12+ / MT4 0.11+; zero on older pods). Event-only P&L accounting is data.net_profit (profit + commission + swap).

Correlate the In and Out events of one round-trip through data.position — on MT5, Out deals carry magic=0 / comment="" unless the closing request set them (platform behavior, not a bridge gap), so position id is the reliable join key. On MT4, deal is always 0 and position equals the order ticket. The same id appears as position in order_history() rows (bridges MT5 0.14+ / MT4 0.13+) and as position_id in position_history().

If the bridge can't fully enrich an event in time it sets data.degraded=True: identifiers, symbol, type, volume and price are still trustworthy, but entry is "Unknown" and the cost fields are zeroed — reconcile that deal via order_history().

async for event in s:
    match event:
        case TradeUpdate() as t if t.data.degraded:
            reconcile_later(t.data.position)     # costs/entry unreliable
        case TradeUpdate() as t if t.data.entry == DealEntry.OUT:
            print(t.data.position, "closed, net", t.data.net_profit)

Errors

Every failure raises a subclass of fxsocket.FxSocketError:

from fxsocket import Client, AccountCapError, InsufficientBalanceError

try:
    fx.accounts.create(server="Demo", login=1, password="…")
except AccountCapError as e:
    print(f"Plan limit reached: {e.current}/{e.cap}")
except InsufficientBalanceError as e:
    print(f"Top up {e.shortfall_eur} EUR first")   # None if the API gave no figure

Private hosting

Dedicated private servers are managed through client.private_servers:

import time

from fxsocket import Client, PrivateAccountStatus, SlotsFullError

with Client(api_key="fxs_live_...", verify_terminal_tls=False) as fx:
    [server] = fx.private_servers.list()
    print(server.name, server.status, f"{server.used_slots}/{server.purchased_slots}")

    try:
        account = fx.private_servers.add_account(
            server, server="ICMarkets-Demo", login=1150125, password="..."
        )
    except SlotsFullError as err:
        print(f"Server full ({err.used}/{err.cap})")
        server = fx.private_servers.resize(server, slots=err.cap + 1)

    # Poll until the on-server agent has the terminal up, then trade as usual.
    while True:
        server = fx.private_servers.get(server)
        account = next(a for a in server.accounts if a.id == account.id)
        if account.status == PrivateAccountStatus.READY:
            break
        time.sleep(5)

    print(fx.terminal(account).account_summary())

Accounts on a private server are traded and streamed exactly like shared-cluster accounts — their rest_url / ws_url simply point at the server's dedicated IP. The server presents a self-signed certificate, so reach it with Client(..., verify_terminal_tls=False) (or supply a pinned CA).

server.cancel_at_period_end is True once a server has been told to stop instead of renewing; it then runs until server.period_end and expires.

Buying, resizing and canceling

fx.private_servers.regions() returns where servers may run, how big they may be and what that costs — call it before buying rather than hardcoding slugs:

options = fx.private_servers.regions()
if options.enabled:
    print(options.region_codes)                     # ['fra1', 'lon1', ...]
    print(options.max_slots, options.max_servers)
    print(options.monthly_price_eur(3))             # Decimal('45') — 3 slots/month

create() buys one, charged to the prepaid balance immediately. It comes back provisioning; poll get() until it is ready, usually a couple of minutes:

from fxsocket import InsufficientBalanceError, PrivateServerStatus, ServerLimitError

try:
    server = fx.private_servers.create(slots=2, region="fra1", name="prop-guard")
except InsufficientBalanceError as err:
    print(f"Top up {err.shortfall_eur} EUR first")
except ServerLimitError:
    print(f"Already own the maximum ({options.max_servers})")

while server.status != PrivateServerStatus.READY:
    time.sleep(10)
    server = fx.private_servers.get(server)

resize() changes the slot count. Increases are prorated over the rest of the period and charged now (the renewal date doesn't move); decreases are free and apply at the next renewal, so paid-for capacity is never destroyed mid-month. Shrinking below the accounts already on the server raises AccountsExceedTargetError — remove accounts first:

server = fx.private_servers.resize(server, slots=4)

cancel() stops the server renewing: it runs until period_end, then expires. resume() undoes that while the period lasts (afterwards the machine is gone and AlreadyLapsedError is raised). delete() destroys the machine and every account on it right away, with no refund for the rest of the prepaid month — prefer cancel() unless you really want it gone now:

server = fx.private_servers.cancel(server)      # stop at period_end
assert server.cancel_at_period_end
server = fx.private_servers.resume(server)      # changed your mind
fx.private_servers.delete(server)               # irreversible, no refund

All of these move the prepaid balance, so they only work on balance-funded servers — a card- or crypto-funded one raises NotBalanceFundedError (a ForbiddenError subclass) and is managed from the dashboard. Read-only fxs_ro_… keys get a plain ForbiddenError.

Wallet

fx.wallet.get() is a read-only view of your prepaid balance: what is in it, top-ups that haven't landed yet, and what the balance will pay for over the next 30 days (account seats and balance-funded private servers together). Amounts are integer EUR cents; the *_eur properties give Decimal euros.

wallet = fx.wallet.get()
print(f"balance {wallet.balance_eur} EUR, covers next 30 days: {wallet.covers_upcoming}")
for charge in wallet.upcoming:
    print(f"  {charge.when:%Y-%m-%d} {charge.kind:6} {charge.label} {charge.amount_eur} EUR")
if not wallet.covers_upcoming:
    print(f"top up at least {wallet.shortfall_eur} EUR")

Affordability is cumulative — with 24 EUR and three 12 EUR renewals the first two are covered and the third is not — so shortfall_eur is the total gap, not the size of any single charge. Topping up happens in the dashboard; the balance is only ever spent through the SDK (account seats and private_servers.create() / resize()), never topped up.

Timestamps

Terminal timestamps (quote.time, candle time, order times) are returned as strings in broker server time — not Python datetime. The trailing Z is stylistic and does not mean UTC. Use terminal.server_timezone() to get the broker's UTC offset if you need to convert.

Requirements

• Python 3.10+
• httpx, pydantic ≥ 2, websockets ≥ 13

Links

• API reference: https://api.fxsocket.com/v1/docs
• Examples: examples/

Development

pip install -e ".[dev]"
ruff check . && mypy && pytest

License

MIT — see LICENSE.