OpenTraderWorld
Your trading workspace, on your own machine.
Alimenter votre agent
Installation de l'éditeur. Ce n'est pas un endpoint PlugMCP.
Host a copy with our keys and logs.
Pas encore de version packagée dans le registre. Suivez le README du dépôt.
URL distante ou commande sur votre machine ? Ce qui distingue les deux installations
Exécuter sur PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
LancerA free, self-hosted platform that bundles the tools traders and investors actually use: trading journal, market data and backtesting, portfolios, news, planning. Your data stays on your machine. And if you want it, your own AI agent runs the whole thing alongside you.
No account, no install, already full of realistic data. It is a shared sandbox, wiped every 15 minutes.
Every few weeks another product promises to be "the Bloomberg Terminal for the rest of us". It is the wrong target, because Bloomberg does not sell software.
OpenTraderWorld is not trying to be the next Bloomberg Terminal. It is the workspace around your own decisions: your journal, your portfolios, your backtests, your research, your automations, enhanced by your favorite AI model. Free, on your machine, with the data providers you already pay for (or the free ones).
You need Docker (macOS, Linux or Windows).
curl -fsSL https://raw.githubusercontent.com/G-OTW/OpenTraderWorld/master/install.sh | bash
The installer asks a few questions, generates strong secrets, pulls the prebuilt images and starts the stack, creates your admin account and prints its password once. Then open http://localhost:5454 and sign in. If you want the long version (manual setup, headless servers, LAN/HTTPS/public exposure, backups), it is all in the documentation.
[!TIP] Prefer to be walked through it? The guided install on the website takes you through it step by step.
If the page won't load over plain HTTP. Browsers with HTTPS-only mode (Chrome/Edge HTTPS-First, Firefox HTTPS-Only, Safari) rewrite
http://intohttps://and then cache that redirect, so a LAN install answers nothing at all. Open the address in a private window to confirm that is what is happening, then allow HTTP for that host. Or switch to the LAN + HTTPS mode in Settings → Network and get a real certificate.
Every image below is the real app, captured from a local install. Prices, news and backtests are real; the journal, portfolio and fundamentals figures are sample data. The images follow your GitHub theme, light or dark.
Workspaces for each hat you wear (investor, trader, quant, portfolio), built from widgets: net worth, positions, live quotes, headlines, upcoming macro releases, goals. Rearrange them or add your own.
Up to twelve linked charts per workspace, drawing tools, built-in studies and alerts. Paste a Pine Script v5/v6 indicator once, keep it in your library, and add it to any chart.
Trade from the chart (experimental): send, move and cancel orders and close positions on Interactive Brokers, Binance, Binance USDⓈ-M Futures and the app's own paper trading account, with your own risk limits and a kill switch. A live account takes orders only once you turn live trading on for it, with its limits set. See Trading from the chart.
[!WARNING] Chart trading is new. It is heavily tested, against simulated venues and the brokers' own test environments, but bugs can remain. Check every order at your broker, start small, and trade at your own risk.
Load a saved strategy, pick a dataset, run. The report puts every trade on the chart, with the figures that matter next to buy and hold. Shown here: an EMA 21/55 crossover on ten years of AAPL daily bars, +375% against +1,217% for buy and hold. The engine does not flatter a strategy.
Your trades with their equity curve, win rate, profit factor, Sharpe and drawdown, filtered by side, ticker, strategy, class or tag. R-multiple analytics, behaviour patterns and a PnL calendar sit one click away.
Holdings priced daily, cost basis, unrealized and realized P/L, allocation by asset or class, and the full ledger of operations behind them.
The risk profile of any stored dataset: volatility, drawdown, value at risk, expected shortfall and the return distribution. Then correlation, efficient frontier, position sizing and derivatives in the next tabs.
Macro series from central banks and statistics offices, charted together with recession shading. Company statements, filings, ETFs, a market calendar and alternative data live in the next tabs.
Your RSS and JSON-API feeds in one stream, polled on a schedule, searchable and filterable by source.
Chain blocks into a workflow: a call to the app's own API or any URL, an AI step, a condition, a notification. Put it on a schedule and read every run afterwards.
Pre-market, in-session and review checklists, each step with its own how-to, tracked by week, month and year.
Settings → Update app tells you when a new version is out. The app cannot update itself (it has no shell or Docker access, by design), so run these on the host. Take a database backup first; migrations run automatically on the new container's first boot.
A default install (prebuilt images, no git checkout) refreshes deploy/ from the latest
release (the public master), then pulls the new images. Your .env, network.env and dns.env are left alone.
cd /path/to/opentraderworld
TMP=$(mktemp -d)
curl -fsSL https://codeload.github.com/G-OTW/OpenTraderWorld/tar.gz/master | tar -xz -C "$TMP"
cp -R "$TMP"/*/deploy/. deploy/
rm -rf "$TMP"
docker compose -f deploy/docker-compose.yml -f deploy/docker-compose.images.yml \
--env-file deploy/.env --env-file deploy/network.env \
pull
docker compose -f deploy/docker-compose.yml -f deploy/docker-compose.images.yml \
--env-file deploy/.env --env-file deploy/network.env \
up -d
A source build (installed with --build, from a full clone) refreshes the checkout and
rebuilds. Releases are fresh snapshots, so reset to the release rather than git pull.
git reset --hard discards any local edits to tracked files.
cd /path/to/OpenTraderWorld
git fetch origin
git reset --hard origin/master
docker compose -f deploy/docker-compose.yml \
--env-file deploy/.env --env-file deploy/network.env \
up -d --build
There is a public sandbox at demo.opentraderworld.com, also linked from the website. No account, no install, and it comes seeded with realistic data so you land in an instance that looks used rather than empty.
One thing to keep in mind: it is shared. Everything you type is visible to other visitors, the whole database is wiped and re-seeded every 15 minutes, connectors are read-only, and settings, the Vault, broker accounts, notification channels and paper trading are disabled. It runs on a small free-tier server with free AI models, so it is slower than a local install. For anything real, install it. Two minutes.
Every module ships with the app. Install what you use, detach the rest.
The full tour, with details, is in the module docs.
One engine, three ways to describe a strategy. No code.
Run any of them on one instrument or a whole portfolio, with exposure limits and a per-asset breakdown.
Take a run and vary its parameters: indicator lengths, thresholds, stops, sizing, costs, portfolio limits, grid settings, DCA amounts, excluded weekdays. Each gets a from / to / step.
Add an in-sample / out-of-sample split and the two stat blocks sit side by side. A wide gap between them is overfitting.
Press one button on a run you like and it keeps going on new candles. There is no second engine and no serialized simulator to drift: each tick reloads the window, runs the ordinary backtest, and diffs the trades against what is already stored. A session opened from a result reproduces that result exactly.
Filter alerts down to the tickers you care about, and pick which channels a session pushes to.
One connector list for the whole app. A connector is a provider account, its credentials (typed in or plugged from the encrypted Vault, write-only once saved), an optional request limit, and the modules allowed to use it. Modules have no provider settings of their own.
Every provider connects directly, with an API key or none at all. Interactive Brokers is the exception: you run IB Gateway or TWS yourself and the connector stores a host and a port. OANDA, TradeStation, FOREX.com and Capital.com are broker feeds: they read the prices of the account you already have there, so they ask for that account's credentials. Details in the connectors guide.
Every timeframe a provider downloads, it also streams, daily and weekly included. A provider that publishes one grain (Alpaca, Massive, Interactive Brokers) has the larger timeframes folded from it, and a session-shaped candle (daily, weekly) is re-read from the provider once it closes.
OANDA, TradeStation and FOREX.com download history but do not stream: their live prices are not served over a WebSocket. A chart can draw the live best bid and ask as two lines, and paper trading fills on them.
Missing a provider? Ask for it and it gets added.
A broker account links to the place you trade, and only reads by default. It imports your fills into the journal, aligns a portfolio on what the account holds, draws positions and working orders on the chart, and reads a tax year. It sends orders only where the broker takes them and you turn Trading on for the account (experimental).
Where a broker issues a view-only key, the form asks for that one. Journal imports from Interactive Brokers go through the Flex account. A live account trades only once you turn Live trading on for it, which requires a size limit per order, a size limit per position and a loss limit. Details in the broker accounts guide.
Macro series, company statements, SEC filings, earnings-call transcripts, ETF holdings, a market calendar and alternative data, read from the sources that publish them and stored in your own database. Every source is a data connector granted to the module; nothing is fetched until you add one. When several providers serve the same data, they are tried in the order you set.
Prices and market ratios come from any market-data connector with daily equity bars. Details in the Fundamentals guide.
These are the notification methods available today. More are planned: if your preferred one is missing, ask for it.
Secrets are sealed on save and never shown again. What can notify you: chart alerts and broker reconcile notices, paper trading sessions, watchlist alerts, reminders, finished or parked data downloads, fundamentals refreshes, new mail, journal, portfolio and tax imports, live feed outages, sign-in alerts, inbound webhooks, and any automator workflow. Everything is off until you switch it on, and in-app notifications work without configuring a single channel.
Push to talk, never always listening: the microphone opens while you hold a shortcut (or click the mic in the top bar) and closes when you release.
Alt+V): what you say becomes a plan. "open journal" works out of the
box in seven languages; anything no command matches can go to the AI agent.Alt+Shift+V): what you say is typed into the focused text field, never
read as a command.Speech engines work like the other brokers: one list in Settings → Voice, keys sealed or plugged from the Vault.
The browser only allows the microphone over HTTPS or on localhost. Details in the voice control guide.
OpenTraderWorld is free and built in the open. If it is useful to you, the easiest way to give back costs nothing:
No pressure. Even just trying it out and telling a friend means a lot. 🙏 Got ideas, or found a bug? See Suggest, vote, contribute below.
This project grows on user feedback:
OpenTraderWorld/
├── install.sh # one-command installer (fetches deploy/, runs setup)
├── deploy/ # docker-compose stack, setup.sh, Caddy config
├── core/ # Rust workspace (API, store, scheduler, feeds…)
├── frontend/ # SvelteKit app (modules UI)
├── docs/ # documentation site (VitePress)
├── scripts/ # validation and parity checks
└── testbook/ # non-regression run against a throwaway instance
Actively developed. It is usable today and moving fast. Check Settings → Update app in the app to know when a new version is out.
This software is for educational and informational purposes only. It is not financial, investment, or trading advice. USE IT AT YOUR OWN RISK. The author and any affiliates accept no responsibility for your trading or investment results, or for any loss or damage arising from use of this software. Never risk money you cannot afford to lose. There will be bugs: the software is provided "as is", without warranty of any kind. You alone are responsible for how you use it and for the decisions you make with it.
FSL-1.1-MIT (Functional Source License). Free for everyone, including commercial use, self-hosting at work, modification, redistribution and forks. The one thing it does not grant is a Competing Use: selling the software, or offering it to others as a paid product or a hosted service. Each release turns into plain MIT two years after it is published.
Free and source-available, self-hostable trading workspace: data, backtest, trading journal, news aggregator, rich-doc editor, portfolio, 20+ modules with a built-in MCP server and AI agent.
Outils
Pas encore de liste d'outils. Les serveurs distants en obtiennent une après le premier contrôle d'état.
FAQ
Copiez sur cette page la configuration stdio pour Cursor, Claude Desktop, Claude Code ou VS Code. Elle utilise l'installation de l'éditeur, pas un proxy PlugMCP.
Aucune variable d'environnement requise n'est indiquée sur cette fiche.
OpenTraderWorld s'exécute en local via stdio.
Dernière vérification : 2026-11-09. Free and source-available, self-hostable trading workspace: data, backtest, trading journal, news aggregator, rich-doc editor, portfolio, 20+ modules with a built-in MCP server and AI agent.