tv-live 0.1.0

Stream OHLCV candles and indicators from TradingView — no Desktop app required

tv-live

Stream OHLCV candles and technical indicators from TradingView — no Desktop app required.

Direct WebSocket connection to data.tradingview.com, the same protocol used by the tradingview.com website in your browser. No Node.js. No CDP. No Electron app running 24/7.

Quick Start

pip install tv-live

# Live stream EURUSD M5 with EMA20 and ATR14
tv-live stream FX:EURUSD 5m --indicators ema20,atr14

# Bar replay — walk through 2000 bars, show signals
tv-live replay FX:EURUSD 5m -n 2000 --quiet

# List available brokers
tv-live brokers forex

How It Works

┌──────────────┐      WebSocket       ┌─────────────────────┐
│   tv-live    │ ◄──────────────────► │ data.tradingview.com │
│  (Python)    │   ~m~ protocol       │   (TV servers)       │
└──────────────┘                      └─────────────────────┘
     │                                        │
     │  Parse timescale_update                │  Same feed as
     │  Calculate indicators locally          │  the website
     ▼                                        ▼
  OHLCV + EMA + ATR + RSI + ...

Unlike CDP-based MCP servers (Jackson, Londarth, tvcontrol) that scrape data from a running TradingView Desktop app via Chrome DevTools Protocol, tv-live connects directly to the same WebSocket servers that power the tradingview.com website. This means:

• No TV Desktop app needed
• No --remote-debugging-port=9222 setup
• No Electron app consuming 300MB+ RAM
• Works on any machine with Python 3.10+

The protocol is TradingView's internal ~m~ format — the same messages your browser receives when you open a chart on tradingview.com. We parse timescale_update messages to extract OHLCV bars, then calculate technical indicators locally using formulas verified to match Pine Script v5.

Commands

stream — Live candle streaming

tv-live stream <SYMBOL> [INTERVAL] [OPTIONS]

Streams candles in real-time. Each completed bar is printed with OHLCV and optional indicator values.

Arguments:

• SYMBOL — Exchange:Pair format (e.g. FX:EURUSD, OANDA:GBPUSD). Shorthand without : defaults to FX broker.
• INTERVAL — 1m, 3m, 5m, 15m, 30m, 1h, 2h, 3h, 4h, 1d, 1w, 1M (default: 5m)

Options:

Flag	Description
-i, --indicators	Comma-separated indicators: ema20,atr14,rsi14
-b, --broker	Broker prefix shortcut (default: FX)

Examples:

tv-live stream FX:EURUSD 5m -i ema20,ema50,atr14
tv-live stream OANDA:GBPUSD 15m -i rsi14
tv-live stream eurusd 1h -i ema200          # shorthand, auto-adds FX:

replay — Bar replay with strategy signals

tv-live replay <SYMBOL> [INTERVAL] [OPTIONS]

Walks through historical bars one by one, calculating indicators cumulatively (only using bars "seen" so far) — the same way a live strategy would. Detects EMA crossover/crossunder signals and prints entry, SL, and TP for each.

Options:

Flag	Description	Default
-n, --bars	Number of bars to fetch	500
-f, --file	Load from CSV instead of WebSocket	—
--ema-fast	Fast EMA period	20
--ema-slow	Slow EMA period	50
--atr	ATR period	14
--sl-mult	SL multiplier × ATR	1.5
--tp-mult	TP multiplier × ATR	4.0
-q, --quiet	Show only signal summary	—

Examples:

tv-live replay FX:EURUSD 5m -n 2000 --quiet
tv-live replay FX:EURUSD 15m --ema-fast 10 --ema-slow 30 --sl-mult 2.0
tv-live replay FX:EURUSD 5m -f historia_EURUSD_5m.csv

Output (--quiet mode):

PODSUMOWANIE SYGNALOW
=====================================================================
  LONG:  4
  SHORT: 4
  CLOSE: 7
  RAZEM: 15

  2026-06-25 01:50:00    SHORT  E=1.13533  SL=1.13554  TP=1.13476
  2026-06-25 03:20:00    LONG   E=1.13659  SL=1.13621  TP=1.13760
  ...

brokers — List available exchanges

tv-live brokers forex     # Forex brokers
tv-live brokers crypto    # Crypto exchanges
tv-live brokers stock     # Stock exchanges
tv-live brokers index     # Index/CFD prefixes
tv-live brokers all       # Everything

info — Configuration reference

tv-live info              # Show supported intervals and indicators

Available Indicators

All indicators use the same formulas as Pine Script v5 — verified by cross-referencing 200 M5 bars against TradingView Desktop (CDP). Close prices match 99.5% of bars; the remaining 0.5% differ by ≤2 ticks.

Indicator	CLI flag	Parameters	Pine Script
EMA	emaN	period=N	ta.ema(close, N)
SMA	smaN	period=N	ta.sma(close, N)
ATR	atrN	period=N	ta.atr(N) — RMA smoothing
RSI	rsiN	period=N	ta.rsi(close, N) — RMA smoothing
MACD	macd	fast=12, slow=26, sig=9	ta.macd(close, 12, 26, 9)
Bollinger Bands	bb	period=20, std=2	ta.bb(close, 20, 2)
VWAP	vwap	—	ta.vwap

Formula accuracy

Indicator	Our formula	Pine Script	Verified difference
EMA(20)	ewm(span=20, adjust=False)	ta.ema(close, 20)	0.00001 (1 tick)
EMA(50)	ewm(span=50, adjust=False)	ta.ema(close, 50)	0.00001 (1 tick)
ATR(14)	rma(tr, 14) = ewm(alpha=1/14)	ta.atr(14)	⚖️ Identical smoothing
RSI(14)	rma(gain, 14) = ewm(alpha=1/14)	ta.rsi(close, 14)	⚖️ Identical smoothing

Data Format

Every bar returned by the stream has the following structure:

{
  "time": "2026-06-26 13:40:00",
  "open": 1.13997,
  "high": 1.14002,
  "low": 1.13982,
  "close": 1.13989,
  "volume": 389,
  "symbol": "FX:EURUSD"
}

This is the same OHLCV format as Jackson MCP's data_get_ohlcv, plus a symbol field. Timestamps are Python datetime objects (convertible to Unix with .timestamp()).

Python API

from tv_live import TvLiveStream

# Basic streaming
stream = TvLiveStream("FX:EURUSD", "5m")

def on_candle(bar, history):
    """Called on every completed candle."""
    print(f"{bar['time']} | O={bar['open']:.5f} C={bar['close']:.5f} V={bar['volume']}")

stream.on_candle = on_candle
stream.on_connect = lambda sym, tf: print(f"Connected: {sym} {tf}")
stream.on_disconnect = lambda reason: print(f"Disconnected: {reason}")

# Start in a background thread (non-blocking)
thread = stream.start_in_thread()

# Later: access data
print(f"Collected {stream.bar_count} bars")
df = stream.df  # pandas DataFrame with all bars

# Stop
stream.stop()

Replay engine

from tv_live.replay import ReplayEngine

engine = ReplayEngine(
    symbol="FX:EURUSD",
    interval="5m",
    ema_fast=20, ema_slow=50,
    atr_period=14,
    sl_mult=1.5, tp_mult=4.0,
)

# From TradingView WebSocket
engine.load_from_websocket(n_bars=2000)

# Or from CSV
engine.load_from_csv("historia_EURUSD_5m.csv")

# Run and get signals
signals = engine.run(verbose=False)
for s in signals:
    print(f"{s['time']} {s['action']} {s.get('entry', s.get('price'))}")

Why This Over CDP-Based MCP Servers?

There are several open-source TradingView MCP servers (Jackson, Londarth, tvcontrol, Ferrox Labs) that connect via Chrome DevTools Protocol to a locally running TradingView Desktop app.

Feature	tv-live	CDP-Based MCP
TV Desktop required	No	Yes (must run 24/7)
RAM usage	~30 MB	~300 MB+ (Electron)
Language	Python	Node.js
Install	pip install tv-live	git clone + npm install + debug port config
Max historical bars	5000+	500 (hardcoded)
Bar replay	Yes (standalone)	Yes (via CDP)
OHLCV data	Yes	Yes
Technical indicators	Yes (local, verified)	Yes (read from TV data window)
Pine Script editing	No	Yes
Strategy tester metrics	No	Yes
Chart screenshots	No	Yes
Pine graphics (lines, labels)	No	Yes
Alert management	No	Yes
Replay mode (visual)	No	Yes

Bottom line: if you need to control TradingView (edit Pine Script, take screenshots, manage alerts, use bar replay visually), use a CDP-based MCP server. If you need data — OHLCV streaming, indicators, bar replay for strategy validation — tv-live gives you the same quality with zero infrastructure overhead.

Limitations

tv-live connects to TradingView's public data feed (unauthorized_user_token). This is the same feed that powers charts for non-logged-in visitors on tradingview.com. As a result:

• No Pine Script access — can't read, write, compile, or debug Pine Script code
• No strategy tester data — can't access backtest metrics, trade lists, or equity curves from TV's Strategy Tester
• No Pine Script graphics — can't read lines, labels, tables, or boxes drawn by custom indicators
• No DOM/order book — can't access depth of market data
• No chart control — can't change chart type, add indicators, or manage layouts
• No screenshots — can't capture chart images

These are all features that require direct interaction with the TradingView Desktop app's internal state — only accessible via CDP.

Project Structure

tv-live/
  tv_live/
    __init__.py        # v0.1.0
    cli.py             # CLI: stream, replay, brokers, info
    stream.py          # WebSocket engine + ping/reconnect
    replay.py          # Bar replay + strategy simulation
    indicators.py      # EMA, SMA, ATR, RSI, MACD, BB, VWAP
    brokers.py         # Broker/Exchange registry (forex, crypto, stock, index)
  setup.py             # pip install
  README.md

License

MIT — do whatever you want with it.

---

Built as an alternative to CDP-based TradingView MCP servers. If you need the extra capabilities (Pine Script, screenshots, etc.), check out LewisWJackson/tradingview-mcp-jackson — it pairs nicely with tv-live for data-only workloads.