Glossary
Definitions of the main terms used in Captain.AI and this manual. As a rule, this manual follows the terminology shown in the product UI.
Core Concepts
| Term | Meaning |
|---|---|
| Job | A definition that bundles the work steps for the AI to execute. The UI consistently uses the term "job". In development documents and the API it is also called an "agent" (AgentGoal), and this manual sometimes writes "job (agent)" |
| Step | An individual execution unit that makes up a job. Steps run in order from top to bottom |
| Worker | The AI execution environment that runs jobs and skills. Administrators register workers, and every job is tied to one of them |
| Skill | An AI work capability that can be added to a worker. Skills are created in the skill development screen (IDE) and deployed to production workers after a request and approval process |
| Chat / Conversation / Session | A series of exchanges with the AI. A conversation that runs a job on a worker is sometimes called a "worker chat", and a conversation that talks directly to an AI model a "lite chat" (regular chat) |
Units of Organization and Sharing
| Term | Meaning |
|---|---|
| Workspace | The top-level container that fully separates data per organization or tenant. Conversations, projects, users, and settings are never shared across workspaces |
| Project | A container that groups related conversations, knowledge, rules, memories, and shared drives. Shared per member or group, with configurable permissions (view only / can use / can edit / admin) |
| Group | A collection of users. Serves as the unit of access control for conversations, projects, workers, jobs, AI models, and more |
| Organization | All users in a workspace. The "entire organization" visibility scope refers to this range |
Information Given to the AI
| Term | Meaning |
|---|---|
| Knowledge (knowledge base / KB) | A collection of documents and data for the AI to reference. There are personal knowledge bases (User Settings) and project knowledge bases |
| Memory | Things the AI remembers across conversations. There are three scopes: personal (my memories) / project / organization |
| Rules | Instructions prepended to every chat within a project |
| Personalization | User settings that teach the AI your role and preferred response style |
| Context (conversation reference) | A feature that passes the content of another conversation to the AI as context for the current conversation |
| Shared drive | A file storage area shared by project members. The chat AI can read and write it directly |
Execution and Operations
| Term | Meaning |
|---|---|
| Deep Research | A research feature where the AI automatically progresses through planning → research → synthesis → review → follow-up research phases and compiles a report |
| Background processing | AI processing that continues on the server side regardless of your connection. It does not stop even if you close the tab |
| Hand off | A feature that lets you continue work in another job or session while keeping the conversation flow and deliverables |
| Credits | A unit representing AI usage. Depending on organization settings, consumption is shown in the header and in chat |
| PAT (personal access token) | An authentication token for accessing the API from external systems. Bound to the workspace it was issued in |
| Preview (early-release) features | Features released ahead of official launch. Administrators can enable them per organization, and their specifications may change or be discontinued without notice |