Skip to content

Factories > Infrastructure

Zero data retention (ZDR) for factory inference

Review provider ZDR coverage, content collection for service improvement, and run-data storage before sending sensitive work to Warp Factories.

Open in ChatGPT ↗
Ask ChatGPT about this page
Open in Claude ↗
Ask Claude about this page
Copied!

Under provider Zero Data Retention (ZDR), AI providers must not retain your prompts and responses. For factory inference, ZDR coverage depends on the model and whose provider account serves the request, not where the worker runs commands.

Warp’s provider ZDR policy applies to supported models on every plan when Warp supplies the provider account. This is called Warp-managed inference.

Factory runs cannot use personal API keys or endpoint configurations stored on your device. For supported alternatives, see inference options for factory runs.

How you check provider ZDR coverage depends on whose provider account serves inference:

  • Warp-managed inference - Avoid models that require provider retention. Check every worker’s model and any separate feature model settings or overrides, not only the foreman’s model.
  • Your provider account - Verify ZDR with your provider for your specific account, model, and configuration. This check applies to team-managed keys and endpoints, BYOLLM, and direct Claude Code or Codex calls. Warp’s agreements don’t establish ZDR for those requests, even if you avoid the listed models. If your endpoint routes to other providers, verify their retention policies too.

Some factory features use models other than the worker’s selected model. Avoiding models that require retention does not by itself confirm ZDR coverage for every AI request. If your organization requires that coverage, contact Warp before sending sensitive work to your factory.

Warp’s content collection for service improvement and storage of factory run data are separate from provider ZDR policies.

Content collection for service improvement

Section titled “Content collection for service improvement”

Team admins manage content collection for service improvement in the Admin Panel’s Privacy settings, subject to plan and workspace restrictions.

The Warp app’s Help improve Warp toggle controls personal analytics and eligible content collection for service improvement. See analytics and storage controls for personal preferences and organization-enforced settings.

Warp stores transcripts, prompt attachments, and uploaded artifacts so you can revisit runs, even with analytics turned off. Destroying a sandbox doesn’t delete these records.

Enterprise admins can configure self-hosted storage separately for transcripts, prompt attachments, and artifacts. Any data type you don’t configure for self-hosted storage remains in Warp-managed storage.

Self-hosting execution does not establish provider ZDR or move saved run data. The Store AI conversations in the cloud toggle controls personal Terminal conversations, not factory run storage.

Reviewing your factory’s inference settings

Section titled “Reviewing your factory’s inference settings”

Review plan eligibility before configuring shared provider keys and endpoints. Team-wide model restrictions in the Admin Panel require the Enterprise plan for Warp Factories.

Review defaults and overrides in the factory web app or your GitHub repository, depending on where your factory definition is hosted: