Skip to content

Core concepts

The vocabulary AgentDock's documentation uses: workspaces, sessions, agents, runs, and tasks.

A handful of terms recur throughout these docs. This page defines them once so later pages can use them without re-explaining.

Workspace

A workspace (also called a project) is a folder on your filesystem that you've opened in AgentDock — just a path plus a display name, stored locally. Sessions belong to a workspace. See Workspace overview.

Session

A session is a single conversation with one agent, scoped to a workspace. Each workspace can hold multiple sessions, and sessions persist between launches. See Session persistence.

Agent

An agent is one of the coding-agent CLIs AgentDock can drive: Claude Code, OpenAI Codex, or Google Antigravity. AgentDock doesn't provide or bundle these — you install and authenticate each one yourself, and AgentDock detects it on your system. See Agents overview.

Run

AgentDock Pro

A run is an AIgency concept: an orchestrated objective made up of one or more tasks, executed by one or more agents, with its own status (draft, ready, running, paused, completed, failed, stopped). See Creating a run.

Task

AgentDock Pro

A task is a single unit of work inside a run, assigned to a specific agent, and optionally dependent on other tasks completing first. See Multi-agent orchestration.