This page defines the core terms of the Agent Hub. It covers the terms you meet on your own computer while you work with the PAS CLI, the terms of the handover to the platform, and the terms you meet in the Control Hub while you deploy and operate what you have built.
The terms are grouped along the path a change travels, from the project on your computer to a running deployment on the platform. For an alphabetical lookup of all terms, see Agent Hub Glossary.
The Product and Its Components
|
Component |
Description |
|---|---|
|
Agent Hub |
The product, in full Scheer PAS Agent Hub. The Agent Hub is an agentic orchestration platform to develop, deploy, run and govern AI agents, workflows and business applications. It covers scaffolding, development, testing, deployment, and operation of software in a single workflow, and it consists of PAS Code and the PAS Cluster.
|
|
PAS Code
|
The local development framework of the Agent Hub. It consists of the PAS CLI and a set of development skills and libraries. You use it to create, develop, and run projects on your computer.
|
|
PAS CLI |
The command line tool that you call through |
|
Control Hub |
The web application in which you deploy and operate the projects that are published to the platform.
|
|
PAS Cluster |
The server side of the Agent Hub. It hosts the Control Hub, the deployment platforms, the platform MCP server, the Agent Gateway, the Git service, and the identity provider. It builds and publishes your projects and runs your deployments. |
|
Platform connection |
The stored connection between your computer and one deployment platform, e.g. dev or prod. You can store more than one connection. Exactly one connection is active, and every PAS CLI command works against the active platform. |
What You Build
You create every unit in this section with the PAS CLI.
|
Term |
Description |
|---|---|
|
Project |
A versioned, deployable unit made of one or more components. The directory that |
|
Component |
A part of a project. A project groups its components so that they are versioned and deployed together. |
|
App |
A component type. A web frontend that users open in the browser. You generate an app with |
|
Service |
A component type. A component that provides functions and data to apps and to other services. You generate a service with |
|
Agent |
An agent is an AI agent that answers questions and performs tasks on the platform. It is not built and deployed on its own: an agent is a feature of a service component, and one service can provide more than one agent. You reach an agent through the service that provides it, so an agent has no endpoint of its own. You generate an agent with |
|
Bundle |
A group of components that are deployed together. A bundle is a package in packages/ and lists its components in a bundle.json, either by name or by glob pattern. You generate a bundle with |
For an overview on all commands and their options, refer to Command Reference.
How You Work
A project runs in three places. The PAS CLI covers the first two.
|
Term |
Description |
|---|---|
|
Local Development |
|
|
Cluster Development |
|
|
Profile |
A named Kubernetes cluster target, stored as a JSON file in |
|
Publishing (to the platform) |
A push builds and publishes your project, and you can deploy it in the Control Hub. |
A local run and a cluster run are not deployments. They exist only on your computer or on your cluster, and they do not appear in the Control Hub.
Working With an AI Coding Assistant
|
Term |
Description |
|---|---|
|
AI Coding Assistant |
The assistant that you use in your editor. Every project that you create with |
|
Skill |
A packaged instruction set that your AI coding assistant loads when it recognizes a matching task. A skill names the PAS CLI command to call, the options that apply, the rules to follow, and what to do with the result.
|
|
Rules |
The code, specification, and documentation rules that the skills apply. A project completes or overrides them in its |
|
Specification |
A |
|
Spec-driven Development |
The fixed order that every change follows: specification, then code, then documentation. Scaffolding a component belongs to the code phase, so an approved specification exists before a generator runs.
|
|
Approval Gate |
The end of a phase. The result is presented to you, and the workflow advances only after you approve the result explicitly. |
|
Platform MCP Server |
The Model Context Protocol (MCP) server of a platform. When you connect your AI coding assistant to it, the assistant gets tools that read the state of the platform.
|
For an overview on all available skills, refer to Skill Reference.
How Your Work Reaches the Platform
|
Term |
Description |
|---|---|
|
Repository |
The Git repository of your project on the platform. |
|
Push |
The handover of your project from your computer to the platform. A push starts the pipeline. |
|
Pipeline |
The platform process that a push starts. The pipeline builds the components of your project, creates a version, publishes the images and the configuration, and registers the version with the platform. The pipeline needs no configuration of its own, because the PAS CLI generated your project to the structure that the pipeline expects. |
|
Version |
The result of publishing a project. Each time a project is published, it becomes a new version. In the Control Hub, you can select the version you want to deploy.
|
|
Projects |
The Control Hub view in which you browse the published projects and their versions, and from which you deploy them.
|
What Runs on the Platform
|
Term |
Description |
||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
(Deployment) Platform |
The target that you deploy a project to. Platforms keep deployments separate from one another, for example a development platform and a production platform. One project can be deployed to several platforms at the same time.
|
||||||||||||||||||
|
Project Deployment |
A running instance of a project on a platform. Each project deployment has its own name and its own configuration, so you can run more than one deployment of the same project.
|
||||||||||||||||||
|
Component Deployment |
The running instance of a single component inside a project deployment. It has its own status, type, endpoint, and configuration.
|
||||||||||||||||||
|
Endpoint |
The address at which a running component can be reached. A component that has no endpoint shows a dash instead of a link, and an agent has no endpoint of its own.
|
||||||||||||||||||
|
Status |
The state of a project deployment or a component deployment:
While a project or component deployment is starting, its indicator is animated. Refer to Control Hub Status and Badge Reference for further information. |
How an Agent Calls a Language Model
A deployed agent calls a large language model (LLM) through the Agent Gateway of its platform. You manage the Agent Gateway in the Control Hub.
|
Term |
Description |
|---|---|
|
Agent Gateway |
The component of a platform through which every call of an agent to an LLM passes. It checks the virtual API Key and the model of the call, and it forwards the call to the LLM provider.
|
|
LLM Provider |
The connection of the Agent Gateway to an account at a company that supplies LLMs. It holds the provider API Key and the default model.
|
|
Virtual API Key |
The key that an agent uses to call LLMs through the Agent Gateway. It belongs to one LLM provider and permits all models, a list of models, or no model of that provider.
|
|
Model Access |
The virtual API Key and the model of a deployed agent. You change it on the agent details page.
|
How a Deployment Is Configured
The configuration has two parts. You create them as files in your project, and an operator customizes them during deployment.
|
Term |
Description |
|---|---|
|
Deployment Configuration |
The runtime settings of a project deployment. You author them as files in your project. They are published with the version and can be adjusted on the deployment afterwards. |
|
Project-level Configuration |
The values that apply to the whole deployment and that its components share. You author them in the configuration file of the project. |
|
Component-level Configuration |
The values of a single component. You author them in the configuration file of that component. |
|
Component Default |
The baseline values of a component, built into its image. A component default applies wherever no other source sets a value. |
|
Reference |
A pointer to a value instead of a copy of the value: to a project-level value by name, or to a secret that is kept in the cluster. The platform resolves a reference when the project is deployed.
|
|
Secret |
A value that is kept in the cluster and used through a reference. A referenced secret appears neither in your sources nor in the published images.
|
|
Precedence |
The order in which a value wins when more than one source sets it. A component default is overridden by a referenced value, and a referenced value is overridden by a value that is set directly in the configuration. |
|
Configuration Editor |
The Control Hub editor in which you view and change the configuration of a deployment. When you apply a change, the deployment reloads the new settings without a full restart.
|
Where You Author the Configuration
Each package has a default location for its configuration file:
|
Package |
Default location |
|---|---|
|
Bundle |
|
|
Service |
|
|
App |
|
The platform resolves a reference, your computer does not. A reference in a file that local development also reads therefore appears as a raw object instead of a value.