Manage your homelab with ChatGPT

A homelab rarely lives on one machine. MapleMCP is designed around multiple enrolled endpoints, so the same authorized AI conversation can move between a workstation, Linux host, server, or other device without pretending they are one undifferentiated computer.

Useful homelab workflows

Inspect before you SSH around

Read a known config, search for a file, inspect process state, or gather system information directly through the selected endpoint's structured tools.

Run commands where the service lives

Use the Processes/terminal capability on the specific enrolled host rather than assuming the AI client's local shell is the target system.

Keep hosts distinct

Device IDs and names stay part of the workflow, which matters when similar services exist on several machines.

Separate permissions by machine

A build box can expose terminal access while another endpoint exposes only files and search. Capability policy does not need to be identical across the lab.

Linux and Proxmox-style hosts

MapleMCP's Linux setup is designed so the AI-facing service runs as a selected non-root Linux user. If installation starts from a root shell on a server or Proxmox host, setup requires an explicit non-root execution user rather than silently falling back to root for ongoing AI operations.

That design does not prevent you from authorizing privileged administration through normal Linux mechanisms. It prevents the remote agent from treating root as an accidental default.

A typical multi-host flow

List enrolled systems.

Start with inventory so the AI knows which endpoints are online and available to the account.

Choose the target explicitly.

Name the host or verify its device ID before a consequential operation.

Gather evidence on that host.

Search files, read configuration, inspect process output, or run a bounded diagnostic.

Make and verify the change.

Run the required command or edit, then check the same endpoint for the expected result.

What MapleMCP is not

MapleMCP does not automatically scan your LAN and take control of every machine it finds. Each endpoint is enrolled and associated with an account. It also does not collapse host identity: a command intended for one server should still be targeted and verified as one server.

Common questions

Can ChatGPT manage multiple homelab servers?

Yes, if the systems are individually enrolled in the same MapleMCP account. The AI can list the available devices and target the one needed for each operation.

Do I need to expose SSH to the internet?

MapleMCP uses its own enrolled-agent and remote MCP path for supported operations. Whether you also expose or use SSH is a separate infrastructure choice.

Can MapleMCP run as root on Linux?

The supported Linux design deliberately runs the AI-facing service as a selected non-root user. Setup from a root shell requires choosing a non-root execution user.

Can different servers have different capabilities?

Yes. Capability groups are selected per device, so one host can expose terminal access while another remains limited to narrower operations.

Related MapleMCP guides

Try MapleMCP on a device you control.

Install the agent, approve the enrollment in your account, and enable only the capability groups you intend to use.

Downloads Review permissions