lupin.Open console
Browsers

Migrate browser workflows without pretending sessions move

Plan a provider cutover around supported contracts, fresh sessions, human reauthentication and verified cleanup.

Lupin editorialAI-assisted3 min read

Migrate the workflow rather than browser memory

A browser-provider cutover should preserve the task's intent and evidence model, not promise that an old live session can move to a new engine. Lupin's current browser product uses Cloudflare Browser Run and stores provider identity with each registered session. Resources from a previous provider can be unavailable to the current path and require a new session. Cookies and sign-ins are not persisted across closed sessions. Begin the migration with an inventory of required actions, account handoffs and outputs so you can rebuild the authorized workflow without depending on an unsupported session-state transfer.

Map only the supported contract

The current page surface supports tabs, public navigation, reading, screenshots and a fixed set of visible-control actions. It disables downloads and arbitrary execution through that surface, and it rejects custom proxy_user_id routing during creation. Map each existing workflow step to an actual supported action and identify gaps before cutover. Do not treat a provider's broader SDK features as Lupin features automatically. For interactive access, the connect action returns private live-view and session-scoped CDP URLs with an access startup deadline. Update client integrations and telemetry redaction around those returned values rather than retaining old connection assumptions.

Prepare the operator and workspace prerequisites

The deployed API needs durable storage and a native BROWSER binding in the correct Cloudflare account. The account needs Browser Run access and available capacity. Workspace keys need the appropriate read, create and delete permissions, along with human verification and billing eligibility. The autonomous bootstrap key cannot allocate paid sessions. Check capabilities and resolve readiness before changing the task runner's production path. Remove obsolete provider configuration only through the operator's controlled change process. A source-reviewed migration plan establishes the intended contract, while deployed provider availability still needs an approved live verification stage.

Illustrative request; source reviewed, not executed against a live provider.
{"name":"provider-cutover-check","timeout_seconds":60}

Cut over with a fresh bounded task

Use a new allocation identity for the intended new-provider session and preserve its Idempotency-Key across retries. Start with a small public navigation, verify the observed URL and content, request access only if needed, and end the session with confirmed cleanup. This is a proposed validation sequence, not a claim that it has already run. For authenticated work, arrange a private human sign-in handoff in the fresh session and verify the intended account afterward. Do not copy old bearer links into the new client or assume that a stored session identifier can recover the previous provider's page state.

Retain rollback evidence and close old resources

Keep task-level inputs, source URLs, completed-step records and verification results outside browser sessions. If cutover validation fails, stop dependent automation and use those records to choose an authorized recovery path. A rollback means selecting a supported operational route, not migrating a closed browser's memory backward. Reconcile any old resources and uncertain allocation or closure outcomes with the operator before creating replacements. Update documentation to state the actual inactivity behavior, finite runtime policy and lack of profile persistence. The migration is complete when the required workflow and cleanup are verified under the new contract, with unsupported assumptions explicitly removed.

Sources & verification

Reviewed against the current Lupin route, provider adapter and shared input contracts, plus the linked official references. Examples were not run against live services.

Prepared with AI assistance by Lupin editorial. How we research and maintain these guides.

Put it into practice

Give your agent the right tools.

Review the API, workspace requirements and permissions before enabling a workflow.

Explore browsers