SCANLINE - TradingView, AI'ed

Three free tiers

What TradingView hands you for free

You do not need a data vendor or a charting startup. TradingView already ships the pieces. Here is the honest map of what you can self-host versus embed.

EMBED // ZERO SETUP

Widgets

Drop-in script tags. TradingView hosts them and supplies the data. Charts, heatmaps, screeners, calendars, news. The whole gallery below is live widgets.

APACHE 2.0 // SELF-HOST

Lightweight Charts

The open-source charting core on npm. Fully redistributable, you bring your own data. This is the real self-host path. There is a live one below.

FREE // ACCESS-GATED

Advanced Charts

The full charting library: drawing tools, indicators, your own datafeed. Free but you apply for access and host it yourself. The pro tier of the DIY path.

YOURS // MCP

The AI layer

A screener plus an MCP server turns all of the above into something an agent can drive: screen, read a chart across timeframes, rank by factor. Pure TradingView.

Live, no account

The widget gallery

Every panel below is an official TradingView widget, embedded as-is. This is the easiest tier: copy a script tag, get a live chart.

Advanced Chartembed-widget-advanced-chart

Symbol Overviewembed-widget-symbol-overview

Technical Gaugeembed-widget-technical-analysis

Stock Heatmapembed-widget-stock-heatmap

Crypto Heatmapembed-widget-crypto-coins-heatmap

Market Overviewembed-widget-market-overview

Screenerembed-widget-screener

Economic Calendarembed-widget-events

Build your own with Lightweight Charts

This panel is not a widget. It is TradingView's open-source Lightweight Charts library, loaded from a CDN and fed data right here in the page. This is the core you can ship in your own product.

Lightweight Chartsnpm i lightweight-charts

The point

Now let your AI drive it

The screener ships an MCP server, so Claude can screen, compare, rank by factor, and read a chart across every timeframe at once. No swapping, no scraping. Pure TradingView data.

# claude_desktop_config.json
{
  "mcpServers": {
    "scanline": { "command": "scanline-mcp" }
  }
}

# then just ask:
"Read NVDA across timeframes and rank
 mega-cap momentum, then chart the leader."

# analyze() answers, multi-timeframe:
NVDA 208.19: Neutral overall.
RSI 46 (bearish); MACD bullish;
timeframes mixed (1w/1m bull, 1h/4h/1d bear).

28 MCP tools and 4 canned prompts. The screening engine, plus symbol intelligence:

analyze chart read, all timeframes screen filters + factors + stats technical_rating the TV gauge compare symbols side by side chart deep link + widget sector_breakdown search_symbols run_factor_preset

Get the screener

Run it yourself

Four ways in, no terminal required for the first

The screener runs on your own machine. There is no hosted version and there is not going to be one: TradingView data is fetched from your own IP, so it cannot live on a cloud box. No account, no API key, nothing to sign up for.

# 1. Download (easiest): no Python, no pip, no Docker
Releases page, then double-click:
  scanline-windows.exe  (SmartScreen: More info, Run anyway)
  scanline-macos        (right-click, Open)
  scanline-linux        (chmod +x, then run)

# 2. Docker
docker run --rm -p 127.0.0.1:8000:8000 ghcr.io/mphinance/scanline

# 3. pipx
pipx install git+https://github.com/mphinance/scanline
scanline           # the web app
scanline-mcp       # the MCP server

# 4. From source
git clone https://github.com/mphinance/scanline
cd scanline
pip install -r requirements.txt
python run.py

# MCP for Claude Code (binary or pipx)
claude mcp add scanline -- /path/to/scanline-linux mcp

# Once published to PyPI: uvx scanline-screener

All four serve the same thing at 127.0.0.1:8000: the full screener, the analytics layer, and the HTTP API. Loopback only unless you say otherwise.

MIT licensed no build step on the frontend no account needed read-only, no order execution

Source on GitHub MCP reference