VectorBT PRO

What it is

The commercial edition of VectorBT, and a separate product from the free one — a separate package (vectorbtpro), a separate repository, a separate price and, by the vendor's own description, "a substantial redesign rather than a drop-in replacement". Both are written by Oleg Polakow. What you buy is membership: collaborator access to a private GitHub repository, the private documentation and release notes, a Discord where the author answers questions, and every release published while the membership is active.

The free card describes what the community edition does. This one is about what the money buys, and the honest short version is that it buys the order model. In the free edition a simulated order is a signal and a market fill. Here the simulator has limit orders with delays, leverage set per order in two modes, contract multipliers, stop ladders, time stops, target prices, negative prices, and callbacks at every stage of the simulation — plus portfolio continuation, which resumes a simulation from saved state instead of re-running the history. Around the simulator sit the things the free edition's own comparison table lists as missing: portfolio optimisers, cross-validation factories, chunking and parallel execution, a CLI, and a Rust backend alongside Numba.

It is still a research engine and nothing else. The vendor's own answer on live trading is that PRO is "the research and signal engine inside your own trading system" — you pull data, run it, take the signals and hand them to an execution layer you wrote. Nothing here connects to an exchange to place an order.

Pricing

Three plans on one page, all visible at once with no billing toggle to mis-read, as of 2026-09-19: $25 a month, $240 for twelve months (shown struck through from $300, which works out at $20 a month), or $500 once for lifetime access. Payment goes through GitHub Sponsors, Ko-fi, Patreon, Stripe, Paddle or Gumroad; crypto is arranged by email. Sales tax may be added depending on where you are.

Each figure is prefixed "Starts at", and that phrasing is doing work: commercial and organisation arrangements are quoted privately by email and no number for them is published. Read the three figures as the individual, non-commercial price.

There is no trial. The vendor says so directly and says why: the open-source edition is the trial. Switching from monthly to lifetime credits half of what you have already paid, with a minimum upgrade payment of $150. Access through GitHub Sponsors is granted immediately; the other payment routes are processed within 24 hours.

The page carries a standing note that this is "a limited-time offer for early contributors" and that the price will rise "to match that of competitors" at each milestone. That has been on the page long enough to be treated as a pricing posture rather than a deadline, but it does mean the figure a reader sees is the figure they lock in.

Stopping payment does not switch the software off. The version installed at that moment keeps running; what ends is the repository, the private site, Discord, support, and every release and fix after that date.

Data & coverage

PRO ships no market data, which is why this card records no latency — but it ships considerably more plumbing to get data in and keep it than the free edition does. The public feature pages add ArcticDB, DuckDB, SQL through SQLAlchemy, PyArrow and FastParquet, HDF, Databento, FinDataPy, TradingView, Polygon.io, Alpha Vantage, Nasdaq Data Link and Alpaca, plus symbol search, synthetic OHLC generation, parallel pulls, and saver classes that keep a CSV or HDF dataset topped up on a schedule.

For crypto specifically, the adapters in the examples are the same two the free edition has: BinanceData and CCXTData, the latter shown pulling BTCUSDT from Binance, Bybit, Bitfinex and KuCoin and merging the four into one frame. That is candle data. CCXT's exchange list is CCXT's, not this vendor's, and its maintenance is CCXT's too.

Integrations

An MCP server, and a real one. vbt mcp serve starts it and vbt mcp search "…" invokes a registered tool straight from the terminal without starting a server at all. Alongside it are function calling across LLM providers, a documentation chat, a search index over the vendor's own knowledge assets, and a source refactorer. This is the concrete difference behind the AI talk in the free edition's README, where all of it is PRO.

The Rust extension ships as platform wheels, so no Rust toolchain is needed to use it from Python; the same simulators are also available as a native crate if that is where you want to work. Three jitting backends — Numba, NumPy and Rust — and seven execution engines from serial through threaded and multiprocess to distributed.

Limitations

The licence is stricter than the marketing reads, and two of the vendor's own documents pull in different directions. The Software Licence grants use "exclusively for Private Use / Non-Commercial Use" and defines non-commercial as use from which "no direct or indirect monetary compensation, commercial advantage, or for-profit purpose is obtained". Organisations get internal R&D only, "provided it does not result in a commercial product, service, or outcome". The separate Remarks page then says licensees "may use, copy, and modify vectorbtpro as long as they do not vend, publish, distribute, or commercialize the source code", and may even ship software that declares vectorbtpro as a dependency so long as end users buy their own licences. Those are not the same rule. Neither document says whether trading your own capital for profit is "Private Use". If that question matters to you, it is an email before it is a purchase.

No funding, no liquidation, no perpetuals — the same wall as the free edition. Leverage is new and genuinely modelled, per order, in two modes. But across all eight public feature pages the words funding and perpetual do not appear once, and the only place liquidation appears is a tutorial where you write a callback that stops the simulation when portfolio value goes negative. There is no maintenance margin, no mark price, no exchange liquidation, and no eight-hourly cash flow. A leveraged perp backtest here still omits the funding leg entirely — it is a better spot backtest with borrowing, not a derivatives simulator. For a crypto reader this is the fact that decides whether $500 buys anything.

Nothing you can read before you buy. The source is in a private repository, the release notes and full documentation are behind the membership, and reverse engineering is prohibited by the licence. The public feature pages are the whole of the public record, and they are explicitly a selection: "only a selection of the most interesting features from each release is highlighted". Which means everything on this card is the vendor's own account of a product nobody outside can inspect.

One person. The copyright, the licence, the support mailbox and the Discord are all Oleg Polakow. That is also true of the free edition, but the free edition's source is public and forkable; here a lifetime purchase is a bet on one author's continued interest, and the card that matters for that bet is the licence's own line that all rights not expressly granted are reserved.

Migration from the free edition is work. The vendor's word is "substantial redesign rather than a drop-in replacement": a separate environment, import vectorbtpro in place of import vectorbt, and adapting existing code against the current documentation. Budget for it rather than assuming a find-and-replace.

No published price above the individual plans. Every tier says "Starts at". If you are buying for a team or for anything commercial, the number on the page is not your number, and there is no public one.

Alternatives

The comparison that matters is not against another vendor, it is against the free edition, and it turns on one question: does your strategy need limit orders, leverage or path-dependent stops? If it does, the free edition cannot express it at any amount of effort and this one can. If it does not — if you are sweeping parameter grids over signals and market fills — the free edition already does that, and does it with the same architecture.

The second question is the licence. If you intend to sell anything built on top, both editions say no by default: the free one through the Commons Clause, this one through a non-commercial grant and a private negotiation. A permissively licensed engine from Backtesting & Research Libraries removes that conversation rather than scheduling it.

And if the strategy is a perpetuals strategy, neither edition is the answer, because neither of them knows what funding is.