Your own charting platform, built by an agent, on your broker's data

A chart that knows your trades

On 20 October, TradingView launched AI Copilot, an assistant that works directly on your charts, as a paid monthly add-on. The same autumn, it also shipped an official agent skill with Lightweight Charts, its free and open-source chart library. That made me wonder whether I could stop paying for charting platforms and instead build a platform of my own, one that knows my trades, runs on my own computer and that I can customise exactly to my preferences.

What I built

Most of what you see on the screen is a copy of what you already know from TradingView or any other advanced trading platform. There are candlestick charts on every timeframe from five seconds up to a month, every symbol my broker offers, moving averages, Bollinger Bands, RSI and MACD, drawing tools, up to four linked charts, alerts and a data export to CSV.

What TradingView doesn't have are the things that come from reading MT5 directly. Every trade I've taken in MT5 is drawn on the chart without me entering anything, with its entry and exit, its risk down to the stop loss, its target and its result in R, the result divided by the risk the trade started with. Next to the chart sits a journal, much like the self-updating trading journal I built earlier, that turns all of those trades into statistics by hour, weekday, symbol and direction, in R rather than in money. The replay plays the market forward candle by candle and shows my own trades at the moment I took them, so I can look at an old trade with only the information I had at the time. And because MT5 keeps the broker's tick data, the platform shows the spread at the moment of every fill and how far each fill was from the price on the order, which you can only measure on your own broker's data.

Why MT5 is the right base

The reason I built it on MT5 is that you're charting on the exact prices your broker quotes you, the ones your orders are filled against, instead of a feed from another source that roughly follows the same market. The spread, the session times and the odd spike are all your broker's, so what you analyse is what you trade, and that's precisely what makes the spread and slippage figures possible. MT5 is also offered by a large number of brokers, so the platform keeps working if you ever move to another broker that offers it, and its Python package reads candles and your trade history from the terminal that's already open and logged in, without your password ever going into the code. TradingView's own data, on the other hand, is off-limits for a project like this, because its policies only allow the data to be displayed, forbid scraping and building products on it, and there's no public API to get it from.

The version we're sharing already covers the boring basics. The platform can only read from MT5, and a test fails as soon as a function that could place, change or close an order appears anywhere in the code, which we checked by having the agent add one on purpose. The page never shows an account number, a balance or an amount of money, it only runs on your own computer, and the bugs that only showed up in a real browser are fixed, including buy arrows that sat just above their candle because MT5 draws candles on the bid price while buys are filled at the ask. TradingView's agent skill kept the agent away from the outdated examples that are all over the web, and the whole build cost about $19 in usage with Claude Opus 5.5.

It costs you nothing to use, it needs no development budget, and we think it can already stand next to other charting platforms when it comes to ease of use. More importantly, anyone can build on it. We've done the majority of the work, and precisely the part that's boring, difficult and heavy on tokens, so you can spend your time refining it and adding the features you want for your own trading, with a minimal amount of work.

The full code is public on GitHub at github.com/purple-technology/hello-purple-charts, where you can read every file before you download anything. You don't need a GitHub account, because the green Code button has a Download ZIP option, and once you've opened the folder with your coding agent, you start from a platform that already works.

What you could add

Once the data, the chart and the safety checks are in place, a new feature is a request to your coding agent rather than a new project. Some of the ideas below are features traders currently pay for in TradingView's higher plans or in dedicated order-flow platforms such as ATAS, Bookmap and Sierra Chart, and others don't exist anywhere, because they only make sense for your own trading and your own broker.

Order flow

Footprint charts split every candle into its price levels and show the buying and selling volume at each one, with the delta and the point of control highlighted. On TradingView they come with the Premium plan and higher, and your agent can build them from the ticks MT5 already provides, drawn as a custom series, a type of plugin TradingView's plugin skill explains. Cumulative delta, the running total of buying minus selling volume, follows from the same data and fits in its own pane.

TPO charts, or market profile, show how much time the market spent at each price during a session. They only need time and price, so they work on every symbol in MT5, forex included, and you could have yesterday's value area and point of control drawn on today's chart automatically, or a separate profile for the Asian, London and New York sessions. A volume profile next to it shows where the most ticks happened.

A read-only DOM shows the buy and sell orders waiting at each price, which MT5 publishes and the Python package can read, with your own entry, stop and target levels marked on the ladder. Record the order book every few seconds and you get a liquidity heatmap in the style of Bookmap. How much depth you see depends on your broker, and our safety test blocks the order book function for now, so adding it is a deliberate choice.

An interactive DOM, where you click a price to place an order, needs permission to trade. That's the moment to go through our safety check for trading tools built with AI, and I'd keep it apart from the read-only chart, so the platform you use for analysis can still never send an order.

Popular chart tools

Smart money concepts fill page after page of TradingView's community scripts, and your agent can draw order blocks, fair value gaps and breaks of structure automatically, using exactly the definitions you trade with instead of someone else's script. Session boxes and kill zones shade the hours you trade, and a multi-timeframe dashboard shows the trend on every timeframe of the symbols you follow in one table. Anchored VWAP from a point you click, seasonality by hour, weekday or month, and a correlation map of the symbols you trade are a single request each.

Backtesting and practice

Manual backtesting, the kind traders pay platforms such as FX Replay for, is a small step from the replay that's already there, because you only need buttons to place practice trades while the market plays forward, on your own broker's data. The replay can also turn into a trade quiz that stops before each of your old entries and asks what you'd do, and a rule-based backtest can run a simple strategy of yours over MT5's history and put its trades in the same journal as your real ones, so you can compare the two.

Risk and funded accounts

A position size calculator turns a stop you draw on the chart into the position size that fits your risk per trade. If you trade a funded account, the daily loss limit and maximum drawdown can appear as price levels on the chart, and an exposure view shows how much risk you have open across all your positions at once, including symbols that tend to move together.

Your own journal

Setup tags give you the journal's statistics per setup, so after a hundred trades you see which of your patterns hold up. A pre-trade checklist has to be filled in before a trade gets its tag, and an automatic screenshot at every entry means your review starts with what you saw at the time. A weekly review can put your best and worst trades, your rule breaks and your execution figures on one page every Friday.

Data from outside MT5

The chart can also show data that MT5 doesn't have, as long as the source allows it. Economic calendar releases can appear as markers on the time axis, COT positioning can sit in its own pane for the markets that have it, index traders could add gamma exposure levels from an options data provider, and crypto traders could add funding rates, open interest and liquidation estimates from an exchange or data service. Keep each connection read-only and check the provider's licence before you build on its data.

An AI that reads your chart

Finally, you could connect a language model that can only read, give it the candles, your trade and your journal notes, and have it describe what happened around a trade you took last week or summarise your week in plain language. Keep the model read-only, and be careful with anything that feeds it outside content such as news, because that's exactly the kind of connection our safety check warns about.

How to start

You'll need Windows with the MT5 desktop terminal open and logged in, and a coding agent such as Claude Code. Download the code from GitHub, open the folder with your agent and let it read the build notes first, because they explain every step and every problem in plain English, so the agent knows how the platform is put together before it changes anything. Then ask for one feature at a time and describe it the way you'd explain it to a colleague, for example a TPO profile for every London session on the 30-minute chart, calculated in the platform's own code, with a test against a small profile you worked out by hand. Check the result against what you see in MT5, run the safety test after every change, and if a feature needs something new from MT5, make sure it's a function that only reads.

There are a few limits to keep in mind. The MT5 Python package only works on Windows with the desktop terminal, alerts only run while the page is open, and older trades have coarser price history. If what you trade isn't available in MT5, the same chart library works with other sources, such as Kraken's free candle data for crypto, but then you're no longer looking at the prices you're filled at.

A first step this week

Pick the one feature from this article, or from your own wish list, that you'd use every day, and write it down in a few sentences the way you'd ask a colleague for it. Then download the code, give those sentences to your coding agent and compare the result with your MT5 terminal. If you don't use MT5, you can try the idea on a smaller scale by asking your agent to draw Kraken's free candle data with Lightweight Charts and its agent skill, and to mark three of your own trades on it.

This article is for information only and is not investment advice. Hello Purple is a technology platform operated by Purple Technology s.r.o. It is not a broker or investment firm and provides no investment service, advice or recommendation. Trading leveraged products involves a high risk of loss. TradingView, Lightweight Charts and MetaTrader are trademarks of their respective owners; Hello Purple is not affiliated with or endorsed by them.

This article is information and commentary. It is not investment advice, not a personal recommendation and not investment research, it takes no account of your circumstances, and it must not be relied on as a reason to trade. Any figure shown is historical or simulated and is not an indication of future results.

All reports →