Skip to main content

Agent System Overview

Users should not have to think about ContextGo as a pile of disconnected runtime concepts.

The better model is:

  • an agent is an execution role
  • a runtime is the execution backend
  • an agent package is a reusable capability bundle
  • skills, hooks, commands, and schedules are layered capabilities

Why this matters

This separation is what lets ContextGo behave like a work system instead of a single-agent wrapper.

It means:

  • the runtime can change without redefining the whole workflow
  • capabilities can be added without rebuilding the assistant from zero
  • the same work surface can support different execution backends