Skip to main content

​
Why versions exist

Every time you save agent configuration — prompt, model, Focuses, linked files, or parameters — XSpada creates a new version. One version is active at a time. That active version is what Chat, Actions, and Tracks use in production. Versions exist because agents are not deterministic. A change that improves one conversation can break another. Version history lets you iterate, compare, and roll back without losing prior work.
Screenshot placeholder Version History tab showing a list of versions with version numbers, dates, and an Active badge on one entry. Include Activate on inactive versions.

​
Active version vs. history

Activating an older version does not delete newer ones. You can move forward and backward as you learn from testing. When an agent is ready for production on Evolve, see Promote an agent to team member.
Treat “Save” as “publish a candidate.” Test in the Lab before activating a version that will affect teammates or automations.

​
The Lab: test before you ship

The Lab tab is a sandbox conversation tied to your agent’s current draft or active configuration. Use it to:
  • Preview replies before teammates rely on the agent
  • Fill in variables with test values via Manage Variables
  • Watch how Focuses and assigned Actions behave in realistic dialogue
  • Catch tone, policy, or tool-selection issues early
Screenshot placeholder Lab tab with chat panel, Manage Variables control, and sidebar of past Lab sessions. Show a test message and agent reply.
Lab sessions appear in the agent’s recent chat list. They are the right place for messy experimentation — not for work you need audited (use Chat Review for that).

​
Lab vs. Chat vs. Action node

“It worked in the Lab” is necessary but not sufficient. Test the same agent in the mode your team will actually use — especially if variables are supplied differently in Actions than in Chat.

​
Studio: build and test together

Studio combines the Builder and Lab in one layout — prompt editor and live conversation side by side. It is the fastest loop for prompt iteration: edit instructions, send a test message, read the reply, edit again. See Agent Studio for when Studio beats switching between tabs.

​
Measuring “good enough”

From What agents are — and aren’t: define what success looks like before you ship. There is rarely one perfect answer to grade against. Start with a handful of representative scenarios your agent must handle — and a short list of things that must never happen.

​
Where to go next

Agent Studio

Combined build-and-test layout.

Chat Review

Audit live conversations after deployment.
If you would like something unique to your business or have a unique issue, use XSpada Evolve or ask a XSpada rep.