Integrating AI into Financial Workflows: Trading Charts, Code Development, and Document Parsing

Financial and data‑analysis workflows often suffer from tool fragmentation: charting, data analysis, and document parsing operate in separate silos. Three open‑source projects address this by providing AI‑driven chart interaction, project‑wide code review, and universal Korean document parsing.

TradingView MCP – AI Directly Operates TradingView Desktop

Technical analysts manually view charts, draw trend lines, measure indicators, and run Pine Script strategies. TradingView MCP bridges Claude Code to a locally running TradingView Desktop via the Chrome DevTools Protocol (CDP), enabling AI to read candlestick data, switch symbols, draw annotations, and compile Pine Script without screenshots.

Chart analysis – AI can query the current chart, retrieve K‑line data, indicator values, and existing annotations.

Pine Script development – AI assists from writing and injecting code to compiling and debugging.

Chart control – AI can change symbols and timeframes, add or remove indicators, and scroll to a specific date.

Drawing annotations – AI can place trend lines, horizontal lines, rectangles, and text labels.

Backtest practice – AI can enter replay mode and step through historical data.

Typical CLI usage:

# Get current quote
 tv quote
# Switch symbol
 tv symbol AAPL
# Capture chart screenshot
 tv screenshot
# Compile Pine Script
 tv pine compile

Installation:

git clone https://github.com/tradesdontlie/tradingview-mcp.git
cd tradingview-mcp
npm install

Repository: https://github.com/tradesdontlie/tradingview-mcp

Oracle – AI Reviews an Entire Project

When a bug cannot be described by a single error message, Oracle automatically discovers relevant source files using glob patterns, packages them in a logical order, and sends the context to large models such as GPT‑5 Pro, Claude Opus, or Gemini 3.1 Pro. The tool works without manual file dragging or API keys by controlling Chrome in a browser mode.

File auto‑discovery – Uses glob patterns (e.g., src/**/*.ts) to locate related files.

Smart packaging – Organizes files in a coherent context rather than simple concatenation.

Multi‑model support – Sends the same prompt to multiple models simultaneously.

Browser mode – Automates Chrome on the ChatGPT page, avoiding the need for an API key.

MCP integration – Available as an MCP tool within Claude Code or Cursor.

Typical CLI usage:

npm install -g @steipete/oracle
# Send project context to AI
oracle -p "帮我 review 项目的架构设计" --file "src/**/*.ts"
# Pipe build error output to AI
make build 2>&1 | oracle -c "帮我看看编译错误"

Repository: https://github.com/steipete/oracle

kordoc – Universal Korean Document Format Parser

Analysts and traders frequently encounter documents in HWP (Korean standard), PDF, XLS, XLSX, and DOCX formats. kordoc is a TypeScript library that parses and converts HWP3, HWP, HWPX, HWPML, PDF, XLS, XLSX, and DOCX, making it suitable for bulk processing of financial statements and reports.

Installation: npm install kordoc Repository: https://github.com/chrisryugj/kordoc