Before you start
- Monk is installed and your agent is signed in. If not, follow the Quickstart. To check, ask
/monk check installation status. - To deploy to a cloud, have access to an account on AWS, Google Cloud, Azure, DigitalOcean or Hetzner. Credentials explains what each one needs. You don’t have to enter anything yet: Monk asks when the plan needs it.
- To try Monk without a cloud account, ask for a local deployment. It runs on your own machine.
1. Ask for a plan
Open your app’s repository in your coding agent and ask:MANIFEST, which describes the services and how they connect. Monk validates that configuration and prepares the plan.
If your project already has a docker-compose file or infrastructure code, the agent uses it to understand the app. Monk doesn’t take over or change your existing deployment. It builds a new one alongside it.
Monk may ask a few questions first, such as which team the project belongs to (your personal account or one of your organizations) and which cloud and region to use.
2. Review the plan in the local dashboard
The plan opens in the local dashboard under Work. It shows:- the services Monk will deploy and how they connect
- the infrastructure it will create, such as the cluster’s provider, region, instance type and count
- warnings and any missing configuration
- the estimated cost
3. Enter credentials
If the plan needs a cloud credential or a secret, such as a database password or an API key for a service, Monk opens a form in the local dashboard. Depending on the provider you type the values, upload a key file or sign in to the provider from the form. Some secret fields can generate a random value for you. Never paste credentials into the chat. The form keeps them away from the coding agent. See Credentials for what each provider needs.4. Deploy
After you approve, Monk creates the cluster, builds the container images and deploys the services in the order theMANIFEST describes. Your agent shows progress in the chat and checks that each service is ready before calling the deployment done.
If something fails, the agent reads the status and logs from Monk, fixes the code or configuration and asks Monk to deploy again. Any new infrastructure or privileged action still needs your approval.
What you get
When the deployment finishes, your agent reports the services that are running, where to reach them and the estimated monthly cost. From then on you can ask:Tear it down safely
When you’re done trying things out, delete the cluster so your provider stops billing for it:If something goes wrong
Next
Deploy an app that's already running
Data moves and traffic cutover
Set up CI/CD
Deploy with GitHub Actions on every push
Watcher and alerts
Crash and resource alerts in Slack
Costs
Estimates, usage and billing alerts

