Getting started
Installing DRADIS
This walks through a first-time install of the AWS Marketplace image, from launching the instance to placing a first real order. Allow about thirty minutes, most of it spent watching the engine trade in ghost mode before you fund anything.
Already running DRADIS and moving to a newer version? Use the upgrade guide instead.
01Launch and secure the instance
Subscribe on the AWS Marketplace listing, then launch from the AMI on t3.medium or larger. Before you open the dashboard, restrict the security group: ports 80 and 443 to your own address only, and port 22 only if you want SSH.
The Control Tower dashboard controls real money once you provide venue credentials. Treat reaching it exactly as you would treat an exchange login, and do not leave it open to the internet.
Name the instance
A one-click Marketplace launch leaves the EC2 Name tag blank, and we cannot set it for you: the instance is created in your account, not ours. Name it yourself at launch, something like dradis-v1.0.4 carrying the version.
This matters most later. Upgrading runs the old and new instances at the same time for a few minutes, and you have to stop the right one. Two blank rows in the console is a bad moment to find that out. The Control Tower footer also shows the running version, so an open tab always identifies itself.
02First login
Open https://<instance-public-ip>/ in a browser. The instance generates its own TLS certificate at first boot, so your browser will warn that the certificate is self-signed. That is expected on a fresh instance.
Sign in with user admin and your EC2 instance ID as the password. The instance ID is a first-run claim token rather than the standing credential: the setup page immediately requires you to create an admin password, which is hashed and becomes the real one. The instance ID stops working at that point.
03Choose your venue
Select Kalshi, Polymarket International or Polymarket US in the setup page. All three binaries ship in the image, but one instance trades one venue at a time.
The change applies when the engine restarts. Restart from the setup page when prompted.
04Enter credentials
Venue API credentials are entered in the setup page and stored on your instance. They are never transmitted anywhere else, and there is no external service in the trading path.
Signal source keys, such as equity market data, are optional and shown separately with a required, recommended or optional tag. A signal source with no key simply idles and publishes a neutral snapshot, and the remaining strategies operate normally. You can add these later.
05Pick a starting profile
Conservative, balanced and aggressive profiles can be applied in one click and then adjusted individually. Conservative is the compiled default and the right starting point.
Applying a profile does not turn ghost mode off. Going live is always a separate, deliberate action.
06Run in ghost mode first
A new install starts in ghost mode, which simulates fills against live order books without placing an order. Leave it on until you have watched the strategies act on real market data and understand why each one is trading or holding.
The dashboard names the specific gate blocking every idle strategy, so inaction is always explained. A strategy sitting still is reporting a reason, not failing.
07Going live, and checking that it took
Turn ghost mode off from the Control Tower when you are ready, and fund the venue account with an amount you are prepared to lose.
Then confirm it actually applied, before you rely on it. Open the Console view and watch a few minutes of output. Every simulated fill is logged with a ghost marker: GHOST_MODE on Polymarket International, and a 👻 line reading ghost entry on Polymarket US and Kalshi. If those keep appearing after you switch, the change has not reached the squadrons already running and you are still paper trading. Restarting does not clear it. Contact support rather than assuming it worked.
How to stop trading in a hurry
If the engine behaves unexpectedly with live capital, stop the instance. That is the action that always works. Standing a squadron down from its page is the next best, and switching ghost mode back on is the least reliable of the three, for the same reason as above: on some versions it does not reach a squadron that is already running.
Resting orders can outlive the engine. Quotes already on the book stay there and can still fill after the process stops, so after an emergency stop always check your venue account for open orders and cancel anything left. From version 1.0.6 the Polymarket International build cancels its own resting orders on a clean shutdown; the Polymarket US and Kalshi builds do not yet, so on those venues this check is the only thing that clears them.
Every strategy can be disabled individually, and every risk parameter is tunable from the browser and applies without a restart.
08Optional: the LLM advisor
The advisor reviews session history and proposes configuration changes with written reasoning. It runs against Anthropic, any OpenAI-compatible endpoint, or a local Ollama model, and defaults to recommend-only. Higher autonomy tiers, which apply changes within bounds you set, are opt-in.
Hosting a model locally
t3.medium cannot run one. A 7B-class quantized model needs roughly 8 GB of RAM for weights alone, so plan on m5.xlarge or larger if you intend to run Ollama on the same instance, and expect CPU inference to take minutes per analysis.
The advisor only calls out when its inputs have actually changed, so a slow local model is workable. A hosted endpoint or a separate Ollama host is the simpler path. Using a hosted provider means paying that provider directly for the tokens the advisor consumes.
09Back up your configuration
Once the instance is configured the way you want it, export a configuration bundle from the Instance Migration panel in the setup page. This is the same bundle used to move to a new version, and it is worth having before you need it.
The bundle contains credentials, including your wallet key. Store it as securely as you would a private key.
If you cannot reach the dashboard
Check the security group first. Ports 80 and 443 must allow your current address, and a changed home IP address is the most common cause of a dashboard that worked yesterday.
If the page loads but the login is rejected, confirm you are using the EC2 instance ID rather than the instance name, and that you have not already set an admin password on this instance. Once set, the instance ID no longer works.
If the engine reports missing credentials after a restart, the venue selected in setup and the credentials entered must match. Selecting Kalshi while holding only Polymarket keys leaves the engine idle by design rather than trading with the wrong venue.
Risk
Trading involves substantial risk of loss. DRADIS executes strategies you choose and configure. It is not investment advice, and no outcome is promised or implied.
Stuck on something this page does not cover? Open a support request or write to [email protected]. Never include private keys, API secrets or seed phrases. No one at DRADIS will ever ask for them.