Configuration
Enabling interviews and selecting a model profile in org Settings.
Interactive Interviews are off by default. An org admin must enable the feature and select a model profile before any project members can run interviews. Configuration lives at Settings → Interview in the Rensei web app.
The Interview settings page is only visible to org admins. It appears in the sidebar only when the screen_interview plan feature is available on the org's current plan.
Plan and feature gate
Interviews require the screen_interview feature to be active on the org's subscription. If your plan does not include it, the Settings → Interview page shows an upgrade prompt instead of the configuration panel.
To check whether the feature is active, open Settings → Interview - if your plan lacks the feature, the page shows the upgrade prompt instead of the configuration panel. (rensei org show prints the org's current plan.) Contact your Rensei account representative to enable it on your plan.
Enabling the feature
Open Settings → Interview in the Rensei web app.
Toggle Enable Interviews to on.
Select a Model Profile from the dropdown. The dropdown lists every model profile that has already been configured for the org. See Model Profiles for how to create profiles.
Click Save. The setting takes effect immediately - no deployment or restart required.
The model profile
The selected model profile is used for every interview run in the org. It determines:
- Which AI model the interview agent uses for its streaming responses.
- Which capacity pool the interview sandbox runs in (the pool is resolved from the profile's configured auth mode and provider).
Auth mode compatibility
Interviews run in the platform's capacity-pool sandbox and support all auth modes:
| Auth mode | Supported |
|---|---|
host-session | Yes |
local | Yes |
cloud (e2b, Vercel, Modal, Daytona) | Yes |
The auth mode is determined by the selected profile. If the profile is configured with host-session, the interview runs on your local worker pool. Cloud auth modes run in the corresponding cloud sandbox.
If the selected profile's capacity pool has no available workers at interview start time, the interview page shows a "Capacity unavailable" message and the participant must retry. This behaves the same as any other agent dispatch against that pool. See Capacity Pools.
Choosing a model for interviews
Interviews are conversational - they benefit from models with strong instruction-following and summarization capability. Recommendations:
- Claude Sonnet - good balance of quality and cost for most teams.
- Claude Opus - higher quality for complex domains or stakeholders who need more depth in questions.
- Avoid lightweight models (Haiku and equivalents) for interviews - the phase logic and MoSCoW elicitation require robust reasoning.
Access control is separate
Enabling the feature at the org level does not automatically grant any project members access to run interviews. Access is controlled per-project. See Access Control for how to grant access.
Disabling the feature
Toggle Enable Interviews back to off in Settings → Interview. This immediately prevents new interviews from starting. In-progress interviews are allowed to complete; they will not be able to submit a new spec after the toggle is turned off.
Configuration reference
| Setting | Where | Default | Notes |
|---|---|---|---|
interview.enabled | Org Settings → Interview | false | Feature flag; requires screen_interview plan feature |
interview.modelProfileId | Org Settings → Interview | null | Must be set before interviews can start; references an org model profile |
| Per-project access | Project Settings → Interview | None granted | Separate from org-level enable; see Access Control |
Related Pages
- Access Control - per-project access grants
- Model Profiles - creating and managing model profiles
- Capacity Pools - how the sandbox resolves worker capacity
- Overview - feature overview and pipeline fit