Chartnaut is a trading terminal that builds itself. You ask an agent for the indicator, the event definition, the study or the app you want, in plain English, and it gets built next to the charts, positions and trades you already have open. Our bet is that this is what every trading terminal ends up being, so we built one now.
Who we built it for
Manual traders. People who read the tape, mark their own levels, take their own trades and want to get better at it. Not developers, and not anyone who wants a bot to trade for them.
If you’ve never written a line of code, you’re exactly who we had in mind. You describe what you want and the agent does the building. There’s a code tab if you ever fancy a look, but you’ll never need it.
We don’t think the manual trader wants agents to take over. They want a second pair of hands for the tedious part, the calculating, the counting, the checking of an idea against history, so the session can go on the decision. So that’s what we built.
Ask for the thing you want
It starts from the question you already have about your market. An agent builds the calculation, marks the event, measures what happened after it, or puts the whole lot in one panel. You look at the result on the chart and change your mind as often as you like.
Indicator agent
The values and levels you read a market by, saved so a definition or study can use them later.
For example
“Plot the London range high and low, then extend both levels to the New York open.”
“Show cumulative delta for the first hour of the cash session.”
Definition agent
A rule for one market event. Check the marks on the chart, then collect every historical match.
For example
“Mark every NQ sweep of the Asian high that closes back below the level within three bars.”
“Find ES opening-range breaks that retest the range high before noon.”
Study agent
A measurement over the events you collected. Rates, breakdowns and the individual rows behind them.
For example
“For these sweep-and-reclaim matches, show how often price reaches the opening range low within 30 minutes.”
“Break the sample down by day of the week and opening direction.”
App agent
One panel with the markets, accounts and research you want to see together.
For example
“Build a desk with ES, NQ and YM, their VWAPs, my open positions and today’s economic calendar.”
“Create an app that lists stocks above the prior day’s high with rising volume.”
Everything stacks
Chartnaut is built in layers, and each one sits on the one below it. An indicator is a calculation on the chart. A definition uses those calculations to say when an event happened. A study takes every time it happened and measures what came next. An app puts any of that in a panel you can open. And anything you can touch in Chartnaut, an agent can touch the same way, so it builds at every layer and hands you the result.
The bottom layers are where we’ve gone furthest. Our indicator and definition scripting is the most advanced we know of. An indicator can read volume at price, a definition can lean on any saved indicator plus price, time and session, and the agent writes both. That’s what makes the agents close to unbounded: if you can describe the level or the event, it can be built.
Studies are the layer nobody else has. A study measures what happens after an event, or after a chain of them, a sweep then a reclaim then a retest, and it keeps working forward. Put a study on the chart and the next time its definition fires, live or in replay, the sample size and the rates are sitting on that bar. It’s a snapshot of a study you ran, so rerun it when you want fresh history.
Apps go anywhere
The top layer is the one we’re most excited about. Chartnaut isn’t only a terminal with an agent in it. The panels themselves are generative, which means you don’t wait for us to ship a feature. You ask for it and it’s yours.
That could be a report on a backtesting session, a screener, or a mini-app for the one market you trade, shared online if you like. Every screen you touch can take a report or an app of your own. Open a trade in the trade detail pane and any app you’ve attached there runs for that trade: a risk app, say, that shows your exposure over the life of the position, one click away every time you open a trade.
Apps can sit in the Terminal, on the dashboard, wherever you want them. We think the scripting and the generative apps are both state of the art, and we plan to keep it that way.
One layout for all of it
Charts is where you build the view. Up to eight panels: charts, a watchlist, depth of market, position management. Save it as a chartbook and it comes back the way you left it, indicators, signal scripts, theme, watchlist and all.
Candlesticks, Heikin-Ashi, footprint, TPO, volume profile, volume bubbles, cumulative delta. Put a profile beside a footprint and a position manager, then use that same chartbook for backtesting, live trading and whatever the agents build for you.
Practise, trade, then look at the record
A backtesting session replays history at the bars and dates you pick, and you place practice trades as it advances. Chartnaut Demo gives you live prices and virtual money. When you’re ready for the real thing, connect an MT4 or MT5 account and trade from the charts you built.
Every closed trade, practice or real, lands in the trade library with its snapshots, notes, tags and playbook. Filter to one setup, one account or one bad week and the Performance page reports on just that. Trade Q&A lets you ask questions of the same record.
A playbook keeps the rules, examples and filed trades for one setup in one place. Once you’ve got a consistent set of labelled examples, train Historical Intelligence on that playbook and it gives you likelihoods for the next one. It’s trained on nothing but your own examples, so the quality of the playbook is the quality of the model.
Think of it as training wheels for your own strategy. Here’s a short video of it in action: you tick what you’re seeing as the trade sets up, and before you commit it tells you how often that exact combination has worked for you.
Bring the AI you already pay for
Plenty of you already pay for Claude Code or Codex, so we made Chartnaut run on them. The desktop app has an AI Providers page (it only exists there) where you pair your subscription. From then on every draft, query and edit an agent makes in Chartnaut goes through that subscription and doesn’t touch your Chartnaut AI credits. This is the way we’d recommend, because you stay in Chartnaut with the chart, the rows and your trades in front of you while the agent works.
If you’d rather start from the other side, the Chartnaut MCP connects from Codex, Claude Code or Cursor. The agent there can read your trades, snapshots, playbooks, backtest sessions, connected accounts and reports, and it builds a study, a query or an app from them. Compare your first-hour fills with the rest of the session. Find the playbook examples that look like the trade you’re reviewing. Whatever it builds opens in Chartnaut, and we wrote up the setup for both routes in its own post.
Before you go
That’s the tour. I’ll finish with a bit about why this exists, since it’s the same reason it’s built the way it is.
I started Chartnaut because the tools I wanted as a manual trader didn’t exist. Everything that came close was built for developers, or for people who wanted the machine to trade for them, and I wanted neither. I wanted to keep making my own calls and have something do the counting, the checking and the building around me.
My bet is that terminals are going to build themselves, and the traders who get the most out of that won’t be the ones who can code. They’ll be the ones who know what they want to see and can say it. That’s who Chartnaut is for, and every product decision starts there.
Next is more of both. Deeper scripting, so the agents can build anything you can describe, and apps in more places, so every screen can carry something you made. All of it on the Claude Code or Codex subscription you already have.
If you try it, start with the market you know best and ask for the one thing you’ve always wanted on the chart. I’d like to hear what it was.
If you want to tell me directly, my inbox is [email protected]. I read it myself and I’ll write back.
Thanks for reading.
Best,
CJFounder, Chartnaut