chartgen

Trading chart generator, MCP server, and live trading engine in Rust. Fetches data from Yahoo Finance and Binance, renders 37 technical indicators as PNG candlestick charts, streams WebSocket quotes into a trading engine with alerts, and integrates with Claude.ai via OAuth 2.1 PKCE.

Quick Start

# Prerequisites: Rust 1.76+, fontconfig/freetype dev libs
sudo apt-get install libfontconfig1-dev libfreetype-dev  # Debian/Ubuntu

git clone https://github.com/5queezer/chartgen.git && cd chartgen
cargo build --release

# Generate a chart
cargo run -- --symbol AAPL --interval 1d -p ema_stack -p vpvr -p rsi -p macd --output chart.png

# Run as MCP server for Claude Desktop
cargo run -- --mcp

# Run as HTTP server for Claude.ai
cargo run -- --serve

# Run in live trading mode (WebSocket feed + alerts + trading tools)
cargo run -- --serve --trade --testnet

CLI

Flag	Description	Default
--symbol / -s	Ticker (AAPL, BTCUSDT)	Sample data
--interval / -i	1m, 5m, 15m, 1h, 4h, 1d, 1wk	4h
-p	Indicator (repeat for multiple)	cipher_b, macd
-n / --bars	Number of bars	120
--width / --height	Image dimensions	1920 x 1080
--output / -o	Output path	chart.png
--source	auto, binance, yahoo	auto
--mcp	MCP stdio mode
--serve	HTTP server mode
--port	Server port	9315
--trade	Enable trading engine (WebSocket feed, orders, alerts)
--testnet	Use Binance testnet (paper trading)

Auto-detection: USDT/BTC/ETH suffixes → Binance, everything else → Yahoo Finance.

Indicators

37 indicators in 5 categories. All configurable via MCP parameters.

Overlays (16) — drawn on price chart

Name	Aliases
ema_stack	ema
bbands	bb, bollinger
keltner	kc
donchian	dc
supertrend	st
sar	psar, parabolic_sar
ichimoku	ichi
heikin_ashi	ha
vwap
vwap_bands
pivot	pivot_points
vpvr	vp, volume_profile
session_vp	svp, session_volume_profile
hvn_lvn	vp_nodes, volume_nodes
naked_poc	npoc
tpo	market_profile

Momentum (10)

Name	Aliases
rsi
macd
stoch	stochastic
cci
williams_r	willr
roc
wavetrend	wt
cipher_b	cipher, cb
adx
rsi_mfi_stoch	rsi_combo, combo

Volatility (3) · Volume (5) · Crypto (4)

Name	Aliases	Category
atr		Volatility
histvol	hv	Volatility
kalman_volume	kalman, kvf	Volatility
obv		Volume
mfi		Volume
cmf		Volume
ad	ad_line	Volume
cvd		Volume
funding	funding_rate	Crypto
oi	open_interest	Crypto
long_short	ls_ratio	Crypto
fear_greed	fng	Crypto

Parameter Examples (MCP)

{"name": "rsi", "length": 21}
{"name": "bbands", "length": 30, "mult": 2.5}
{"name": "vpvr", "bins": 32, "side": "right"}

MCP Integration

Claude Desktop

{
  "mcpServers": {
    "chartgen": {
      "command": "/path/to/chartgen",
      "args": ["--mcp"]
    }
  }
}

Claude.ai

Add https://chartgen.vasudev.xyz as a remote MCP connector. OAuth 2.1 PKCE is handled automatically.

Web frontend

A SolidJS browser SPA lives in web/ — one-screen chart viewer with the same OAuth PKCE flow. See the Web frontend guide.

Tools

Charting:

• generate_chart — render a chart. format can be png (default), summary (compact JSON with OHLCV stats, last indicator values, signals, divergences, and levels — 10–50× cheaper in tokens than a PNG), or both.
• list_indicators — discover all 37 indicators with parameters.
• get_indicators — compute indicators and return raw numeric values as JSON.

Trading (require --trade):

• place_order, cancel_order, get_orders — order lifecycle
• get_positions, get_balance — account state
• set_alert, list_alerts, cancel_alert — price / indicator-signal alerts
• get_notifications — drain triggered-alert queue (poll)
• subscribe_notifications, unsubscribe_notifications, list_subscriptions — push notifications via SSE

Architecture

src/
├── main.rs              CLI + mode dispatch
├── lib.rs               Library root
├── data.rs              Bar, OhlcvData
├── indicator.rs         Indicator trait, PanelResult, Line/Fill/HBar
├── indicator_state.rs   Streaming indicator state for the engine
├── mtf.rs               Multi-timeframe aggregation
├── indicators/
│   ├── mod.rs           Registry (by_name, available, registry)
│   ├── ta_bridge.rs     ta-rs wrappers
│   ├── custom.rs        custom indicators (incl. volume-profile family)
│   ├── cipher_b.rs      Cipher B (Pine Script port)
│   ├── external.rs      API-based indicators
│   ├── macd.rs          MACD
│   ├── rsi.rs           RSI
│   ├── rsi_combo.rs     RSI + MFI + Stoch combo panel
│   └── wavetrend.rs     WaveTrend
├── fetch.rs             Yahoo Finance + Binance REST
├── feed.rs              Binance WebSocket kline feed
├── engine.rs            Trading engine (live bars, indicators, alerts)
├── trading/
│   ├── alert.rs         Alert conditions and matching
│   ├── exchange.rs      Exchange client abstraction
│   ├── order.rs         Order types + store
│   ├── position.rs      Position tracking + PnL
│   ├── signals.rs       Shared signal-label constants
│   ├── subscription.rs  Push-notification SubscriptionRegistry
│   └── persistence.rs   Alert/order persistence
├── renderer.rs          plotters-based chart rendering
├── mcp.rs               MCP JSON-RPC (stdio + HTTP shared)
└── server.rs            axum HTTP, OAuth 2.1 PKCE, SSE

Deployment

Docker

docker build -t chartgen .
docker run -p 9315:9315 -e CHARTGEN_BASE_URL=https://your-domain.com chartgen

2-stage build: rust:1.86-slim → debian:bookworm-slim (~93MB).

CI/CD

Push to master → GitHub Actions runs check/test/clippy/fmt → on success, Coolify auto-deploys via its /deploy endpoint.

Required secrets: COOLIFY_API_TOKEN, COOLIFY_APP_UUID, COOLIFY_BASE_URL.

Testing

cargo test          # e2e + unit tests
cargo fmt --check   # enforced by pre-commit hook
cargo clippy -- -D warnings

License

MIT