Choose your next step

GroundControl is installed. What next?

Go from a claimed instance to one understood workload and a verified agent connection.

YOU WILL FINISH WITH

You know your active host, have chosen an existing or new application, and can verify the next step.

On this page

First, know what installation completed

The launcher starts GroundControl privately, creates persistent storage and connects the local Docker workspace or destination Linux server. Claim establishes your administrator account. You do not need to add that same host again using SSH.

After claim, desktop onboarding verifies private local access. On a Linux server, you can publish a stable HTTPS origin for remote MCP clients or keep it private while evaluating. On completion, GroundControl probes the host and shows its findings.

Already doneStill your decision
GroundControl is running and claimedWhich existing applications to enroll, or which new application to deploy
The local host is connectedWhether to add another remote server
Onboarding can inspect host stateWhether the detected paths and proxy match your actual setup
OAuth and MCP are built into the instanceWhich client gets access, to which workloads and capabilities

Review the onboarding scan

The scan looks for the operating system, Docker, containers, reverse proxy and existing application folders. It can ask you to clarify an unknown proxy, a missing proxy configuration path, containers without Compose labels, or a Compose project without running containers.

Answer from your real setup. A detected file or route is an observation, not a health result. If your app lives outside the configured discovery paths, it may appear in the broader onboarding findings but be missing from the deployment inventory. Adjust Settings → Layout and rescan.

Check before continuing

You recognize the active server and can identify where one application lives. If the probe failed, resolve the host connection before treating an empty list as an empty server.

Choose your first application

Bring one existing workload into view before granting an agent access.

0 / 5 checked
  1. Guide
  2. Guide
  3. Guide
  4. Guide
  5. Guide

Your checklist for this page session. Checking a box does not run an action or verify a server.

Start in Projects → Add project and pick a folder or GitHub repository. In that project, choose Add deploymentto connect an existing workload or launch a new instance. Theprojects guide explains each path.

Know where to go in the product

WorkspaceUse it for
OverviewOrientation to the selected host, its condition and the applications running there.
DeploymentsScan candidates, enroll a workload, and open its Manage, Deploy, Environment and Configure controls.
ProjectsAdd an application from a folder or GitHub and organize its deployment instances by host and environment.
RuntimeInspect containers and their state when you need the lower-level picture.
TemplatesChoose, validate, configure, review and deploy a new workload.
Intelligence / AlertsFollow failures and investigate the runtime and public-route evidence.
AgentsCopy the MCP endpoint and review or revoke authorized clients.
Settings → VPS / LayoutManage host connections and confirm where GroundControl looks for workloads.
TerminalUse the human operator shell for work that requires direct host access.

Finish with a small, observable success

Open your enrolled application, confirm its source and containers, and record the real public URL. Connect your agent after you can identify that workload. Start with a read-only request:

First request to your connected agent
Use GroundControl to list my available deployments. Inspect the one I select and check its runtime and public health. Show what you verified and what is still unknown. Do not change anything.
Check before continuing

The agent calls GroundControl, sees only granted deployments, and returns actual tool evidence. A public check may be skipped when no public URL exists; that is not proof that the customer endpoint works.

Move to a controlled redeploy when the identity and recovery plan are understood. Enable merge automation only after its managed-mode and repository requirements are met.