A local MCP bridge and CLI that lets an AI assistant read, control, and automate a locally running TradingView Desktop application via the Chrome DevTools Protocol. It enables LLM-assisted chart analysis and Pine Script development without executing real trades.
Project overview
Gives developers a way to connect AI agents directly to TradingView Desktop for chart interaction, visual analysis, and script automation through local orchestration.
Project type
MCP · AI Coding · Workflow
Deployment
Refer to project documentation
License
License pending
Best for
Developers and researchers needing to automate local TradingView Desktop chart interactions and analysis via an AI agent.
Key capabilities
Write, inject, compile, debug, and iterate on TradingView Pine Scripts with AI assistance.
Change symbols, timeframes, zoom to dates, and add or remove indicators.
Read chart indicator values, price levels, and annotations for AI visual analysis.
Draw trend lines, horizontal lines, rectangles, and text annotations on charts.
Create, list, and delete price alerts on TradingView.
Step through historical bars and practice entries and exits.
Capture chart state screenshots for AI visual analysis.
Set up multi-chart grids such as 2x2 or 3x1 with different symbols per pane.
Limitations and risks
Relies on undocumented internal TradingView APIs which can break without notice.
Cannot execute real trades and is restricted to chart interaction only.
Will not work if TradingView changes their internal Electron structure.
Programmatic consumption of TradingView data may conflict with their Terms of Use.
Users may face account bans, suspensions, or legal actions from TradingView.
Getting started
Setup involves cloning the repository, installing dependencies via npm, launching TradingView Desktop with a debugging port, adding the MCP server to Claude Code config, and verifying the connection with a health check.
Alternatives and comparisons
Provides tested open-source AI agents and agent skills, including agents that connect to external tools via Model Context Protocol.