Expected outcome
Your team and Tempera have a small evaluation plan with an agreed workflow, data boundary, owners, and success criteria.
You can use this public guide without a Tempera account, product credential, or private workspace. Product access begins only after private onboarding.
Before you start
- One staging or test workflow
- An owner for the agent integration
- A person who can review failed or sensitive runs
Steps
- 01
Choose one workflow
Bring a browser task with a known failure mode or review boundary. Avoid starting with an entire agent program.
- 02
Set the data boundary
Agree which staging data, browser state, screenshots, and trace fields may be used. Identify anything that must be redacted or excluded before a session runs.
- 03
Choose each record you need
Decide which result matters for the evaluation: a Tempo session and human decision, a Palette trace and review annotation, a repeatable test case, or a blocked staging release. Do not treat these as one automatic record.
- 04
Receive environment-specific access
Tempera provides the workspace, connection details, and approved package or client configuration for the evaluation. If the evaluation uses both Tempo and Palette, onboarding also defines the explicit correlation field and how your team will check it; correlation is not automatic. This guide does not publish a shared API endpoint or public package install command.
What success looks like
- One named staging workflow
- A written data and review boundary
- An integration owner and a review owner
- A named correlation field and check when both Tempo and Palette are in scope
- Two or three measurable evaluation outcomes
If something is blocked
- If the evaluation needs several products at once, narrow it to the first browser session and the evidence required to review it.
- If production data is required on day one, pause and complete the security and retention review before connecting it.