What to use instead of Gekko

The product

Gekko was the crypto trading bot most people met first. It was free, MIT-licensed, and it ran on your own machine — a Node.js process, a SQLite file of candles, and a web interface you opened on localhost. In that interface you imported history from an exchange, backtested a strategy against it, watched the equity curve, ran the same strategy in what the docs call "fake money" mode, and then, when you were ready, pointed it at a real key. One tool, one screen, the whole loop. It shipped a set of technical-analysis strategies and more than 130 indicators, and a strategy was a JavaScript file with three callbacks — init, update(candle) and check(candle) — that ended in this.advice('long') or this.advice('short'). You could read one in five minutes and write one in an afternoon. That is why it collected 10,177 stars and around 3,800 forks, and why every "how to build a trading bot" post of the 2017 cycle ended up pointing at it.

Its author stopped. The repository was archived by the owner on 16 February 2020 and is now read-only, and the README leads with the sentence "I am officially not maintaining this project anymore." The commit that added that sentence, on 16 February 2020, is the last one on the develop branch. The last commit that changed the program rather than the notice was on 29 April 2019 — so the honest end date for the software is earlier than the end date for the project.

What makes Gekko worth a page in 2026 is that it does not look dead. Checked on 21 September 2026, the documentation site answers with HTTP 200 and so does the forum. The docs even tell you the truth: "This project is not maintained anymore" sits at the top of every page. Underneath it, the installation page still instructs you to run git clone git://github.com/askmike/gekko.git -b stable, and GitHub permanently disabled the unencrypted git protocol on 15 March 2022 — a clone with that URL now hangs and times out rather than failing with anything a reader could interpret. The prerequisites underneath say Node "at least version 8.11.2". The dependency list in package.json pins koa at 1.x, lodash at 2.x and sqlite3 at 4.x. And the supported-exchanges page still lists GDAX, BTCC, wex.nz and Quadriga CX among the venues you can trade on.

The forum is the same shape of trap. It is up, it accepts registrations, and its busiest sections — Strategy Development, Technical Support — have last posts dated 2021. A search engine will happily hand you a thread about your exact error. The thread is five years old and describes an exchange API that has been replaced twice since.

What to use instead

Expect to rewrite the strategy, not to migrate it. Nothing in this catalogue reads a Gekko strategy file, and nothing reads its SQLite candle database either. What transfers is the idea — the indicator, the threshold, the rule you convinced yourself of — and the order of this list is how much work stands between you and that idea running again.

If what you want back is the browser tab. OctoBot is the closest thing here to the experience of running Gekko: you install it, open a web interface, connect exchange keys and configure a strategy in the UI rather than in a file. Grid, DCA, basket and TradingView-signal strategies are configured rather than coded, and it backtests, paper-trades and trades live on 20 tested exchanges. The difference that will bite is that OctoBot's custom strategies are Python packages, not a file you paste into a text box — so if your Gekko use was picking a bundled strategy and turning two knobs, this is a short trip, and if it was writing your own, it is not.

If what you want back is the whole loop, in code. Freqtrade is the practical successor and the one most Gekko refugees end up on. Free, GPL-3.0, self-hosted, and it covers the same arc in one tool — it downloads the candles itself, backtests, hyper-optimises, dry-runs against live prices, then trades live with the same strategy class. The cost of entry is Python and a different mental model, which the migration notes below get into.

If the backtest was the point. Plenty of people never let Gekko near a real key and used it as a research tool. Jesse is built for that reading: a Python research framework with a much better-specified engine, spot and perpetuals, and a candle importer. Two caveats worth knowing before you install it. Its live and paper trading run through a closed-source plugin that costs $899 one-time at the Basic tier, where Gekko's live mode was free; and its free tier limits live trading to exchange testnets. If the plan was always to go live for nothing, Freqtrade is the answer and Jesse is not.

If you were maintaining the exchange layer anyway. A fair number of Gekko installations were kept alive by someone patching exchanges/ to keep up with a venue's API changes. That job now has a maintained home: CCXT is one MIT interface over 104 exchanges, 76 of them over websocket, in eight languages including JavaScript — which is the one pick here that lets you stay in Node. It is not a bot. It gives you fetching, normalisation and order entry, and the strategy loop, the scheduler, the state and the backtester are yours to write.

Two things this list deliberately does not contain. Hummingbot is excellent and is for market making, which is not what Gekko did. And no fork of Gekko is on it, because none of them has a card here and recommending an unexamined fork of an abandoned codebase to somebody who is about to give it an API key is not a recommendation.

What you give up

JavaScript, and the shelf of strategies written in it. This is the real loss. Gekko's strategy format was small enough that a whole cottage industry of free and paid .js strategies grew around it, and none of them run anywhere else. Three of the four picks are Python. CCXT speaks JavaScript but has no strategy format to speak it in. Whatever you had — your own file, something from the forum, something you bought — is now a specification you re-implement, not an asset you move.

The operational weight of one process and one file. Gekko was npm install, node gekko --ui, and a SQLite database you could open with any tool on your machine. Freqtrade's recommended path is Docker Compose with a config JSON, a strategies directory and a data directory. Jesse expects a database and a cache service alongside it. OctoBot ships desktop builds and is the lightest of the three, but none of them is a single Node process and a file you can sqlite3 into. If Gekko was running on a Raspberry Pi in a cupboard, budget for that changing.

Ten years of search results that are about your exact problem and are now wrong. A project with 10,000 stars leaves a very deep trail — forum threads, Stack Overflow answers, YouTube walkthroughs, blog posts. All of it is still indexed, none of it is marked as expired, and the docs and forum being live is what keeps it ranking. The replacements have better documentation and a fraction of the folklore, so the first time you hit something odd you will be reading a reference page instead of finding somebody who had the same error in 2018.

Migration notes

The strategy's shape changes, not just its language. Gekko called check(candle) once per candle and you kept your own state between calls — a currentTrend variable, a counter, a flag. Freqtrade hands your strategy a whole pandas DataFrame and asks you to populate entry and exit columns across it at once. Code that was written as a state machine does not translate line by line into a vectorised transform, and the place it silently goes wrong is lookahead: in Gekko you could not see the future because you only had one candle, and in a DataFrame you can, trivially, by reading a row you have not reached yet. Check every shift and every rolling window.

• Your candle database is exportable, and its timestamps are in seconds. Gekko's SQLite candles table is start, open, high, low, close, vwp, volume, trades, and start is written as candle.start.unix() — unix seconds. Every engine here wants milliseconds. A history imported without multiplying by 1,000 lands in January 1970 and backtests as an empty range, not as an error. The vwp column — Gekko's volume-weighted price for the bar — has no equivalent anywhere on this list; drop it.
• Do not port the pair list. Four of the venues on Gekko's supported-exchanges page no longer operate, and the ones that do have renamed markets since. Rebuild the list from the exchange's current symbols rather than from your old config.
• Fees and slippage were a config value, and the default is not the same anywhere else. Gekko's paper trader took a fee percentage and a slippage percentage and filled at the candle close. Each engine here ships its own defaults and its own fill rule, and the difference is worth more than most strategy tweaks — what a crypto backtest silently assumes is the version of this argument with the engines named.
• Gekko was spot-only. If the plan on the other side is perpetuals, the cost model gains funding payments and a liquidation check, neither of which your old results ever accounted for. A strategy that looked flat on spot can look very different once funding is settled every eight hours.
• Check who is paying for the free bot you land on. Gekko took nothing and pointed at a donation address. Several of the maintained free bots are funded by the venue instead, through builder fees, broker codes or referral links — who pays for your free trading bot lists the five mechanisms and where to find them.
• Anything dated after February 2020 that claims to be a Gekko release is not from this project. The repository is read-only. A package, installer or "Gekko Pro" build offered elsewhere is somebody else's code under a name people trust, and it wants your exchange key.

The rest of the field is in Crypto Trading Bots.