Core concepts
| Term | Definition |
|---|---|
| Agent | A configured AI specialist in your workspace — instructions, model, Focuses, and optional Actions — used in Chat, Actions, and Tracks. |
| Agent Manager | Hub for My Agents, Chat, Chat Review, Studio, and Artifacts. |
| Model | The AI reasoning engine behind an agent. Fixed at training time; you choose which model powers each agent. |
| System prompt | Standing instructions: role, rules, tone, and boundaries. Applies every turn. Set in the Builder tab. |
| User message | What someone asks in this turn — in Chat, Lab, or input from an Action node. |
| Variable | Placeholder in the system prompt filled at runtime (e.g. customer name, date range). |
| Focus | Capability package granting tools and domain access — Contact Book, Media, builders, Actions Focus, etc. |
| Assigned Action | A specific Action pinned to an agent; triggerable during conversation. |
| Actions Focus | Focus that lets an agent search and run Actions dynamically without pre-assigning each one. |
| Skill | Importable procedural playbook (instruction module) assigned to an agent — how to run a specific workflow. |
| Skills library | Workspace catalog of imported skills available to assign to agents. |
| Team member agent | A promoted agent with a seat on your organization team — can work on Evolve tasks and requests. |
| Promote (agent) | Link a finalized agent to a team member record so it appears on My Team and can take Evolve work. |
| Linked context | Files permanently attached to an agent configuration; included in every run. |
| Chat-linked file | File attached to one conversation only. |
| Artifact | Structured output an agent creates (document, table, component, etc.) — saved and versioned. |
| Builder Focus | Focus that unlocks an artifact type (text editor, table editor, image studio, etc.). |
| Version | Saved snapshot of full agent configuration after a save. |
| Active version | The version used for live Chat, Action nodes, and Track work. |
| Lab | Sandbox chat on the agent detail page for testing configuration. |
| Studio | Combined Builder + Lab layout for fast prompt iteration. |
| Chat | Production conversation with an agent’s active configuration. |
| Chat Review | Workspace-wide history of agent conversations — audit and continue threads. |
| Context usage | Indicator of how much of the model’s context window is consumed. |
| Harness | The scaffolding around the model — loop, memory, tools, guardrails, logging. XSpada provides much of this; you configure the rest. |
Mental model terms
| Term | Definition |
|---|---|
| Rails | Traditional automation — fixed steps, same path every time (Actions often play this role). |
| Driver | Agentic approach — goal-seeking, adaptive, not guaranteed. |
| Loop | Observe context → decide → act → repeat until the goal is met. |
| Prompt engineering | Wording instructions so the model behaves as intended. |
| Context engineering | Choosing what information the model sees each turn. |
| Tool engineering | Designing what actions the agent can take (Focuses, Actions). |
Artifact types
| Type | One-line purpose |
|---|---|
| Text editor | Rich documents and drafts |
| Table / spreadsheet editor | Structured tables and data grids |
| Component builder | Form and UI layouts |
| Code component | Editable code |
| Image studio | Generated or edited images |
| 3D model viewer | 3D asset viewing |
| Sprite sheet | Sprite and animation layouts |
| Registry component builder | Components for Intelligent Registries |
| Query config tester | Interactive registry query testing |
| Live web preview | Runnable web previews |
Chat modes compared
| Mode | Purpose |
|---|---|
| Lab | Test while building |
| Chat | Day-to-day work |
| Use Agent node | One-shot run inside an Action |
| Chat Review | Audit past conversations |
When to use what
| I want to… | Start here |
|---|---|
| Understand what agents are (and aren’t) | What agents are — and aren’t |
| Write instructions and variables | System prompt & variables |
| Give an agent tools | Focus & capabilities |
| Attach files or get artifacts | Context, files & artifacts |
| Test and version safely | Versions & testing |
| Create my first agent | Create an agent |
| Walk every configuration tab | Configure an agent |
| Iterate prompt + chat together | Agent Studio |
| Talk to an agent | Chat with an agent |
| Audit conversations | Chat Review |
| Browse saved outputs | Working with artifacts |
| Put an agent in an Action | Agents in Actions → Agent node |
| Put an agent on a Track | Agents in Tracks |
| Import skills for an agent | Agent skills |
| Promote an agent to the team | Promote an agent to team member |
| Explore fit for my business | Use Case Brainstorm agent in Evolve |
Example agents by job
| Job | Suggested Focuses | One-line purpose |
|---|---|---|
| Contact researcher | Contact Book, Chat | Answer questions about people records |
| Policy assistant | Generic + linked context PDFs | Answer from internal policies only |
| Drafting aide | Text editor, Chat | Produce editable documents from briefs |
| Pipeline classifier | Generic (Use Agent in Actions) | Label and route incoming text in automation |
| Track overseer | Contact Book + assigned Actions | Advance onboarding instances, fulfill tasks |
For deeper examples tailored to your industry, use the Use Case Brainstorm agent in Evolve.