---
title: Run Freqtrade in the Cloud
description: Move a local freqtrade bot to a managed runtime without changing the strategy.
---

You have a freqtrade strategy that works — and a laptop that sleeps, a home IP the exchange throttles, or a VPS you'd rather not babysit. This guide moves that exact bot to [Superior Trade](/quickstart/introduction)'s managed runtime. The strategy file does not change.

Two scope facts up front, so you don't read ten minutes to learn them: freqtrade on Superior currently targets **Hyperliquid**; Binance is coming soon. If your bot trades Kraken, Bybit, or another exchange, the runtime can't host it yet (the [venue matrix](/context/venues) is the live list). And the runtime executes **live/paper strategies**; `hyperopt` and FreqAI training jobs aren't part of the deployment surface today — validate and tune locally, deploy the result. Pricing: a Free tier exists (3 deployments); see [Plans](/account/usage).

## What carries over unchanged

- **Your strategy class.** The same `IStrategy` subclass, same `populate_indicators` / `populate_entry_trend` / `populate_exit_trend`, same imports (`talib.abstract as ta`, `qtpylib`).
- **Your config keys.** `timeframe`, `stake_amount`, `max_open_trades`, `pair_whitelist`, ROI/stoploss tables — passed through in `config`.
- **Freqtrade itself.** The runtime runs the official freqtrade image at a [pinned version](/runtime/frameworks), not a fork. Behavior you tested locally is behavior you get.

## What Superior replaces

| Local problem                              | Runtime answer                                                                          |
| ------------------------------------------ | --------------------------------------------------------------------------------------- |
| Machine must stay awake                    | Managed process, restarts on failure, `health.restarts` visible                         |
| Residential IP throttling / geo-blocks     | Live and paper deployments run in venue-proximate regions with clean egress             |
| `config.json` holds your keys in plaintext | Managed wallet credentials today; [BYOK injection](/runtime/credentials) is coming soon |
| Logs on a box you have to SSH into         | [`GET …/logs`](/runtime/deployments) over the API                                       |
| "Is it still running?"                     | `GET /runtime/deployments/:id` from anywhere                                            |

## The move, step by step

**1. Backtest on our data first** — same code, [`POST /runtime/backtests`](/runtime/backtests). This validates that the runtime's data and your local results agree before anything runs live.

**2. Paper-deploy** — [`"mode": "paper"`](/runtime/paper-trading) with your real config. Let it run a few days; compare its entries to your local dry-run.

**3. Use the managed wallet** — the default [managed trading wallet](/wallet/overview) needs no key handling at all: deposit USDC and skip to step 4. Bringing your own account with [BYOK](/runtime/credentials) is coming soon.

**4. Go live** — `"mode": "live"`, then [`PUT /runtime/deployments/:id/status`](/runtime/deployments) with `{ "action": "start" }`. Set `alive_until` for a bounded first run — a bot that stops itself on schedule is a feature, not a failure. Check `status`, tail logs, read [metrics](/runtime/deployments#get-runtime-deployments-id-metrics).

**5. Retire the local bot** once fills line up. Or don't — nothing here locks you in; the same file still runs anywhere freqtrade does.

## Differences to expect

- Configure the public exchange name and pairs under `config.exchange`; the runtime injects only the credential-backed connection details. Never put secrets in that block — it will be rejected as a [credential-in-config](/runtime/credentials) violation.
- Copy the `framework_symbols.freqtrade` form from [`/context/markets`](/context/markets) into both runtime backtest `symbols` and `config.exchange.pair_whitelist` (e.g. `BTC/USDC:USDC` on Hyperliquid). If `symbols` is omitted, the runtime derives it from `pair_whitelist`; if both are present, they must match exactly.
- Managed **backtest runners deny all network egress** and read their market data from the mounted Superior dataset. Custom pairlist handlers and external data fetches therefore cannot run during a managed backtest.
- Live and paper deployments are a separate workload: gVisor isolates them and blocks cluster-internal and private networks, while general outbound HTTPS is currently permitted. External dependencies remain unsupported and may be restricted in a future hardening pass; a third-party outage can still fail a deployment at runtime.