How to Self-Host Amp Orbs on Your Own Infrastructure

Amp Code reported that developers and infrastructure teams can now self-host orbs within their own environments, providing the necessary compute while Amp handles the transformation into operational machines. The setup allows systems with fast machine initialization and pause-and-resume capabilities to scale down to zero, removing the need to maintain a large pool of idle active servers. Organizations implement this capability by running an Amp orb provider within their infrastructure, which connects directly to Amp to receive requests for spawning, pausing, or deleting orbs.

Deploying Custom Infrastructure Dispatch Programs

How these operational requests are satisfied depends entirely on the host environment, as the provider dispatches commands to a user-provided program. Amp Code provided illustrative dispatch examples covering Docker, E2B, Kubernetes, Apple containers, Modal, Vercel, Daytona, and exe.dev. For example, Docker setups use environment variables and shell scripts to start, stop, or remove containers based on the incoming orb identifier and operation type. Similar custom scripts manage cloud sandboxes across platforms like E2B and Modal by creating dedicated container templates and handling file snapshots during pauses.

Configuring Kubernetes Clusters and Workspace Tokens

Teams can also use Amp’s Helm chart to run orbs on Google Kubernetes Engine, Amazon Elastic Kubernetes Service, or Azure Kubernetes Service. This method requires a dedicated cluster with worker nodes and snapshot storage so paused orbs can save state and resume on alternative nodes. Administrators create a custom orb provider in their Amp workspace, copy the workspace token, and execute the provider script with the assigned token. Once running, users access Amp interfaces, expand the orb menu, select their self-hosted provider, and submit prompts.

Standard Orb Pricing Schedule for Enterprises

Self-hosted orbs remain free for all individual and team workspaces until then.