TradingView MCP: Give Your AI Assistant Market Data, Safely
By Shah Rukh, software developer · · 8 min read
TradingView MCP is a small server that hands your AI assistant a set of market-data and technical-analysis tools. Once it is connected, you can ask Claude, ChatGPT, Cursor or Copilot things like “show today’s top crypto gainers on Binance” or “run a technical analysis of NVDA”, and the assistant fetches real numbers instead of guessing. This is how to run it free on your own machine, plus the paid hosted option, and the one caution I’d want a friend to read first.
Read this before anything else. This server produces market data for research and learning. It is not financial advice, and nothing it prints, no “BUY”, “SELL”, score or “trade setup”, is a recommendation to buy or sell anything. The project says this plainly, and so do I: the data can be delayed, incomplete or just wrong, markets lose people money, and any decision you make is yours alone. Treat every output as a starting point for your own research, not a signal to act on.
What it actually is
MCP stands for Model Context Protocol. It is a standard way to plug extra tools into an AI assistant. On its own, a chatbot has no idea what Bitcoin did this morning. An MCP server fills that gap: it exposes a list of functions the assistant can call, and this one is all about markets.
The README describes it as a TradingView MCP server for real-time market data, technical indicators, screeners and backtesting, covering stocks, crypto, forex and futures. It bundles 37 tools in total, things like a full technical-analysis read (RSI, MACD, Bollinger Bands and more), multi-exchange screeners, a backtesting engine with nine strategies, and a market snapshot of major indices. You ask in plain language; the assistant picks the right tool.
One thing worth clearing up, because the name confuses people: despite “TradingView” in the title, this is an independent open-source project, not affiliated with or endorsed by TradingView Inc. The README is explicit about that. It also does not log into, scrape or automate a TradingView account. It needs no TradingView account or API key at all; it pulls data server-side from public endpoints. That is different from the other kind of “TradingView MCP” that remote-controls the TradingView desktop app, so don’t mix the two up.
Who it’s really for
If you’re a self-taught developer or a curious beginner who already chats with an AI assistant and wants it to reason over live numbers, this is a tidy way in. It’s also handy if you’re learning how indicators and backtests work and want to poke at them conversationally. If you just want price alerts on your phone, this is overkill. And if you’re hoping for something that tells you what to buy, please re-read the warning above, because that is exactly what it is not.
What you need
- Python 3.10 to 3.13. The project supports
>=3.10,<3.14. Python 3.14 is not supported yet and causes a specific failure I’ll cover below. - UV, a fast Python package manager the project uses to install and run everything.
- An MCP client, the thing you’ll talk to. The docs walk through Claude Desktop and Codex, but any MCP client works.
- An internet connection, since all the data comes from third-party sources.
- No API keys are required for the core tools. One optional free key (Marketaux) unlocks news and sentiment, covered later.
The licence is MIT (copyright Ahmet Taner Atila), which is about as permissive as it gets: use it, change it, ship it, just keep the licence notice. The software comes with no warranty, which for a market-data tool is worth taking literally.
A few real precautions
None of these are scary, but they’re the honest fine print:
- Only install from the official sources. The README carries a security notice: this project is Python-only and has no official npm or Node.js package. Copies published elsewhere have added a
package.jsonthat installs malware through npm. The genuine sources are the GitHub repoatilaahmettaner/tradingview-mcp, the PyPI packagetradingview-mcp-server, and the hosted server. If you ever cloned a “tradingview-mcp” from another account and rannpm install, the project says to treat that machine as compromised. - Know where your secrets live. The server holds no credentials of its own. The only secrets are optional tokens read from environment variables or a local
.envfile (the Marketaux news key and optional proxy details). That.envis git-ignored so it won’t get committed. If you do set a token, keep it out of screenshots and shared configs. A tool like our password generator is handy for any secret you need to create. - Data goes to third parties. Your symbol lookups hit upstream providers (TradingView’s data, Yahoo Finance, and Marketaux if you enable news). That’s normal for a data tool, but it means your queries aren’t entirely private.
- It’s informational, full stop. The SECURITY file itself says to treat tool output as informational, not as trading advice. I keep repeating this because it’s the whole point.
Self-hosting on Mac
This is the free route. First install UV. On a Mac the simplest way is Homebrew:
brew install uv
No Homebrew? Use the official script instead:
curl -LsSf https://astral.sh/uv/install.sh | sh
Check it worked:
uv --version
Now point your MCP client at the package. For Claude Desktop, the config file lives at:
~/Library/Application Support/Claude/claude_desktop_config.json
One Mac gotcha the README flags: GUI apps like Claude Desktop sometimes can’t find uvx on their PATH, so use the full path to it. Add this block (swap in your username):
{
"mcpServers": {
"tradingview": {
"command": "/Users/YOUR_USERNAME/.local/bin/uvx",
"args": ["--from", "tradingview-mcp-server", "tradingview-mcp"]
}
}
}
Quit Claude Desktop completely and reopen it. The first launch downloads the package and builds a fresh environment, so give it a minute; the docs note first startup can take 30 to 60 seconds. Then ask it: “Can you show me the available TradingView tools?” If you see names like top_gainers and bollinger_scan, you’re set.
Self-hosting on Windows
Install UV from PowerShell:
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
Open your Claude Desktop config. Press Win + R, type %APPDATA%\Claude, press Enter, and create or edit claude_desktop_config.json there. Here’s where I’d save you some grief: on Windows the project documents a real bug. On first launch you can hit MCP error -32001: Request timed out, and the cause is Python 3.14, which isn’t supported yet. The fix is to pin Python 3.13, which has prebuilt wheels:
{
"mcpServers": {
"tradingview": {
"command": "uvx",
"args": ["--python", "3.13", "--from", "tradingview-mcp-server", "tradingview-mcp"]
}
}
}
On Windows uvx usually lives at %USERPROFILE%\.local\bin\uvx.exe if the bare command isn’t found. If UV itself isn’t recognised after install, the docs suggest adding C:\Users\YourUsername\.local\bin to your PATH and restarting the terminal. Restart Claude Desktop, wait for it to warm up, and test with the same “show me the tools” question.
Optional: news and sentiment
Two tools, financial_news and market_sentiment, plus the news parts of the “power tool” combined_analysis, are powered by Marketaux. Without a token they return a friendly “not configured” note and everything else keeps working. If you want them, grab a free Marketaux key (100 requests a day, and the server caches for hours so that goes further than it sounds) and set it:
export MARKETAUX_API_TOKEN=your_token_here
There’s also an optional proxy setting for getting around rate limits on some sources; you can ignore it entirely and the core TradingView tools still work.
First real use
Once it’s connected, the README’s starter prompts are a good warm-up: “Run a full technical analysis of NVDA”, “What’s the multi-timeframe read on gold?”, or “Backtest an RSI strategy on BTC on the daily timeframe.” The install guide suggests starting on the KuCoin exchange for crypto and using standard timeframes like 15m, 1h and 1D, which it says are the most reliable. If you want to keep a record of a screener’s results, you can paste the table out and convert it with our CSV to JSON tool, or tidy up a write-up with Markdown to HTML.
The paid hosted option
Don’t want to touch Python at all? The project also offers a hosted version at pro.cryptosieve.com, which serves the same 37 tools as a single connector URL you paste into Claude, ChatGPT, Copilot or Cursor. The project prices it at $9/month for the Pro tier, with a 3-day free trial (there’s also a higher-limit Pro+ tier at $29/month). The project is clear that self-hosting stays free forever and the hosted version is just for people who’d rather skip running and updating it themselves. I’d only reach for it if the ops genuinely aren’t worth your time; the free route isn’t hard.
Things that can go sideways
- “No data found.” The docs suggest trying a different exchange (KuCoin tends to work best), using standard symbols like
BTCUSDT, and checking your timeframe format. - Slow first request. Normal. The first call warms things up and can take 10 to 30 seconds; later ones are quicker. Asking for 5 to 10 results instead of 50 helps too.
- Claude can’t see the server. Usually a JSON typo in the config or an incomplete restart. Check the syntax and fully quit and reopen the app.
- Uninstalling. The repo doesn’t document a dedicated uninstall step. Removing the server’s block from your
claude_desktop_config.jsonstops the client from loading it, which is enough for most people.
I wrote this from the project’s repo as it was on October 5, 2026 (version 0.9.0, latest commit dated October 4, 2026). Details like prices, tool counts and supported Python versions can change, so check the project’s own README and the hosted site before you rely on them. ToolsCloset and ToolsCloset are not connected to, affiliated with, or endorsed by this project or TradingView Inc.; this is an independent guide. And once more, gently: this is market data for research, not financial advice. You trade at your own risk.
Frequently asked questions
Is the TradingView MCP server free?
Yes. Self-hosting it on your own computer is free under the MIT licence. There is also a paid hosted version at pro.cryptosieve.com, which the project prices at $9 a month for the Pro tier with a 3-day free trial, for people who would rather not run it themselves.
Do I need a TradingView account or API key?
No. The README says the server needs no TradingView account and no TradingView API key, and it does not log into or automate a TradingView session. The core tools work with no keys at all. One optional free key (Marketaux) unlocks the news and sentiment tools, and everything else still works without it.
Is this an official TradingView product?
No. It is an independent open-source project, not affiliated with, endorsed by, or associated with TradingView Inc. “TradingView” is a trademark of its owner. The project simply consumes public third-party market data.
Can I use its output to decide what to buy or sell?
No. The project states clearly that nothing it produces is financial advice, and that its signals, scores and “trade setups” are not recommendations to buy, sell or hold anything. Treat the output as research data only, do your own homework, and if you need advice talk to a licensed professional. You trade at your own risk.
Why does it time out on Windows the first time?
The project documents this. It happens with Python 3.14, which is not supported yet. The fix is to pin Python 3.13 in your config by adding "--python", "3.13" to the args, since 3.13 has prebuilt wheels. After that the server starts normally.
Is it safe to install?
The official Python package is the one to use. The project warns that it has no official npm or Node.js package, and that copies published under other accounts have added a package.json that installs malware through npm. Stick to the GitHub repo atilaahmettaner/tradingview-mcp, the PyPI package tradingview-mcp-server, or the hosted site, and never run npm install on a copy from somewhere else.
Where does it store my secrets?
The server keeps no credentials of its own. The only secrets are optional tokens (the Marketaux news key and optional proxy details) that it reads from environment variables or a local .env file, which is git-ignored so it will not be committed. If you do not set those, there are no secrets to store.
More AI tips
Free Claude Code Setup Guide
Run the Claude Code app with free AI models on Windows or Mac. Full steps, plus the precautions most guides skip.
EmailOSINT Review and Guide
See which accounts, data breaches and stolen-password logs are tied to your email, and what to do about each result.
FastMCP for beginners
Turn a plain Python function into a tool an AI assistant can call. Setup on Windows and Mac, step by step.
Set up Context Mode the easy way
Context Mode is an MCP plugin that stops your coding agent's context window from filling up with raw tool output.
AWS MCP servers, explained
AWS publishes a big set of open-source MCP servers. Here's what they do and how to wire one into your AI coding assistant.
MCP servers in TypeScript
A beginner-friendly look at mcp-use for building and connecting MCP servers, clients and agents in TypeScript.
OpenHands Agent Canvas Guide
Set up the OpenHands control centre for AI coding agents on Windows or Mac, with the safety steps to take first.
Dify Self-Hosting Guide
Run the Dify AI app builder on your own computer with Docker Compose, and know the costs and licence limits first.
LibreChat Docker Setup Guide
Run your own ChatGPT-style chat app on your computer with Docker. Full steps, real costs and safety checks.
Unity MCP Setup Guide
Let Claude Code, Cursor or Copilot work inside the Unity Editor. Full install steps, plus the precautions to take first.
sprite-gen Setup Guide
Turn one character drawing into a transparent game sprite sheet. Setup steps, real costs and precautions.
Logo Design Skill Setup Guide
Add a free logo-design skill to Claude or another AI agent on Windows or Mac, with the precautions to take first.
Openvid Screen Demo Guide
Record your screen and turn it into a polished demo in the browser, or run Openvid yourself on Windows or Mac.
ReelMimic Setup Guide
Set up ReelMimic to make an original 2D animation in the style of a reference video, with the copyright checks to do first.