TradingView MCP: Market Data and Technical Analysis for AI Assistants via MCP

What TradingView MCP Does and Who It Serves

AI assistants connected via the Model Context Protocol can call tools in external servers during a conversation. TradingView MCP is one such server: it exposes market data and analysis capabilities as MCP tools that Claude, ChatGPT, Cursor, GitHub Copilot, and other MCP clients can invoke directly when a user asks a market question. The README describes four representative prompts: asking for the top crypto gainers on Binance, running a full technical analysis of NVDA, getting a multi-timeframe read on gold, and backtesting an RSI strategy on BTC. These become straightforward requests rather than custom Python scripts. The project targets developers who want programmatic market data access, and traders who want to use an AI assistant for analysis without switching between a browser and a chat window. The README notes clearly that the project is not affiliated with, endorsed by, or associated with TradingView Inc., and that 'TradingView' is a trademark of its respective owner.

The 37 Tools and What They Cover

The README describes the server as providing 37 technical analysis tools. The May 2026 update notes list several categories of tools: `yahoo_price` and `stock_extended_hours` for price data, `top_gainers` and `volume_breakout_scanner` for screeners, `multi_timeframe_analysis` and `combined_analysis` for compound analysis, and `financial_news` for news. The backtesting suite includes nine strategies as of that update: the original six plus `rsi_pullback`, `keltner_breakout`, and `triple_ema`. The `compare_strategies` tool ranks all nine. The README describes a resilience layer added in May 2026 with automatic retry and a 60-second TTL cache on the TradingView screener provider, reducing transient errors on the compound analysis tools. For financial news, the README describes a rebuild that replaced deprecated Reuters RSS endpoints with Yahoo Finance, MarketWatch, and CNBC. Optional Marketaux integration is available when `MARKETAUX_API_TOKEN` is set in the .env file.

Installing and Running the Self-Hosted Server

Install the package with pip:

pip install tradingview-mcp-server

Or run with Docker:

docker-compose up -d

The docker-compose.yml maps the container's port 8000 to host port 8080 and sets `PORT=8000` and `HOST=0.0.0.0`. The Dockerfile builds in a python:3.11-slim image with uv as the installer and produces a runtime image that runs the server as a non-root user. The entry point is:

tradingview-mcp streamable-http --host 0.0.0.0 --port 8000

Environment configuration uses a .env file. The .env.example documents two optional dependencies: the Marketaux API token for financial news and a Webshare proxy for Yahoo Finance and Xueqiu access behind rate limits. The `PROXY_ENABLED` variable defaults to true in the example but can be set to false to run without proxy. The pyproject.toml constrains the Python version to `>=3.10,<3.14` and pins `mcp[cli]>=1.14.0,<2` to avoid a breaking change in the 2.x SDK release.

Connecting to Claude and Other MCP Clients

The README lists the hosted endpoint at pro.cryptosieve.com as the zero-setup option: paste the connector URL into Claude.ai, ChatGPT, Copilot, or Cursor and all 37 tools become available without installing Python or running a server. The hosted service costs from nine dollars per month for the Pro tier or twenty-nine for Pro+, with a three-day free trial. The README compares this to the self-hosted path, which stays free indefinitely but requires managing the Python environment. For Claude.ai specifically, the README describes an MCP Apps feature that enables the server to render an interactive candlestick chart (with daily, hourly, and fifteen-minute views, and an optional Bollinger Band overlay) directly inside the conversation. The README notes this renders 'live inside the conversation via MCP Apps' and that 'the AI reads the band values back to you.'

How This Differs from Desktop-Automation TradingView MCPs

The README contains an explicit comparison section addressing a naming ambiguity. Two kinds of projects now use the name 'TradingView MCP.' Desktop-automation MCPs use Chrome DevTools to remote-control TradingView Desktop and require a logged-in TradingView session, usually with a paid TradingView plan. They work best in Claude Code on a local machine and stop functioning when the machine is off. TradingView MCP (this project) fetches market data from public endpoints without a TradingView account, no browser automation, and no logged-in session. The README states: 'This server does not log into, scrape, or automate a TradingView session.' The trade-off is that this project cannot interact with the TradingView UI directly: it cannot draw on charts, edit Pine Script, or use TradingView's replay feature. For those use cases, the README points to a separate repository.

Limitations and What the README Flags

The README carries a prominent disclaimer: nothing produced by the server is investment, financial, legal, tax, or accounting advice. Outputs including indicators, scores, signals, trade setups, entries, stop losses, and targets are computed from third-party market data and are explicitly described as not being recommendations to buy, sell, or hold any asset. Market data may be delayed, inaccurate, or incomplete. The news tools require an optional `MARKETAUX_API_TOKEN` to function; without it, the README states the news tools return empty results while all other tools work normally. The proxy configuration is optional but the README notes that without it, Yahoo Finance and Xueqiu tools may hit rate limits. The mcp package is pinned below version 2.0 because the 2.x release removed the `mcp.server.fastmcp` module the server imports, and moving to the new SDK is described as a deliberate migration rather than a drive-by update.

Licence and Maintenance

TradingView MCP is licensed under MIT. The last push to the repository was on 2026-09-01. The repository has no GitHub releases; versioning is managed through the pyproject.toml version field, which shows 0.9.0. The project includes a CHANGELOG.md, CONTRIBUTING.md, SECURITY.md, and a EXAMPLES.md with worked examples. The repository structure separates source code into the src/ directory, tests in tests/, and deployment files (Dockerfile, docker-compose.yml) at the root. The project is available on PyPI as tradingview-mcp-server. The README notes GitHub sponsorship is available for contributors who want to support ongoing development. The pyproject.toml specifies Python 3.10 through 3.13 (excluding 3.14) and pins the mcp package to versions 1.14.0 through less than 2.0. The upper bound on the mcp package version is intentional: the 2.x release removed the mcp.server.fastmcp module that the server imports. The pyproject.toml comment explains this directly, noting that moving to the 2.x SDK is a deliberate migration rather than an automatic upgrade. Teams building on this server should be aware of this constraint when managing dependency updates in their environment. One practical difference between the two deployment modes is uptime: the README's comparison table notes that the hosted endpoint keeps answering 24/7, while the self-hosted server stops responding when the local machine is off. The docs/ directory in the repository contains additional documentation beyond the README, and the openclaw/ directory holds configuration for OpenClaw, an alternative MCP registry integration.

Editorial conclusion

TradingView MCP is the right tool for developers and traders who want to query market data, run screeners, or backtest strategies from inside an AI assistant conversation without writing a custom API integration. The self-hosted version is free and covers all 37 tools; the hosted endpoint at pro.cryptosieve.com removes the Python setup at a monthly cost. The README's disclaimer is worth reading carefully: nothing this server produces is financial advice, and its outputs are computed from third-party data that may be delayed, inaccurate, or incomplete. Verify the Python version constraint (3.10 to 3.13, excluding 3.14) and the optional `MARKETAUX_API_TOKEN` requirement before deploying for the financial news tools.

Frequently asked questions

Official sources

1. atilaahmettaner/tradingview-mcp on GitHub
2. Issues
3. License: MIT
4. Project website
5. README