The agent detail page
Each agent opens to a detail view with tabs across the top. Together they cover the four layers from Introduction: identity, instructions, brain, and capabilities.
Save from the top right after meaningful changes. Saving creates a new version — see Versions & testing.
Overview
Overview is the at-a-glance dashboard: description, active version number, model, parameters, assigned Focuses, and context usage. Recent Lab and Chat threads appear here for quick return visits. Edit the description inline with the pencil control. Deeper edits use other tabs.Screenshot placeholder
Overview tab with active version card, model name, context usage meter, and recent chats section.
Focus
Assign Focuses — capability packages that grant tools and domain access. An agent can hold multiple Focuses; start with the minimum set and expand after Lab testing. See Focus & capabilities for the full Focus catalog and assigned Actions vs. Actions Focus.Model Select
Browse models authorized for your workspace. Recommended models appear first; remove the recommended filter to see the full catalog. Consider speed vs. quality vs. context window size for your use case. After changing the model, retest in the Lab — the same prompt can behave differently across models.Builder
The heart of configuration. Write or refine the system prompt, define variables, open Settings for parameters, and use helpers (Guided setup, Edit with AI, XSpada Intelligence). See System prompt & variables for prompt engineering guidance.Actions
Assign specific Actions this agent may trigger during conversation — a curated toolkit. Distinct from Actions Focus, which lets the agent discover Actions dynamically.Skills
Assign skills — imported playbooks that teach procedural workflows without bloating the system prompt. Import skills at the workspace level, then attach them per agent. See Agent skills.Linked Context
Link files from your workspace that should always inform this agent. Contents are injected into every run. Keep the list tight — linked files consume context budget. See Context, files & artifacts.Version History
Every save adds a version. One is active for live use. Activate any older version to roll back. Compare behavior by activating a version and testing in the Lab.Lab
Test conversations without affecting production expectations. Use Manage Variables when the prompt includes placeholders. Lab history appears in Overview and in the Lab sidebar.Switch to Studio
For heavy prompt iteration, open Studio from the agent sub-nav — Builder and Lab side by side. See Agent Studio.Where to go next
Agent Studio
Combined build-and-test layout.
Chat with an agent
Use the configured agent in day-to-day work.