FAQ
The questions teams ask before they put memory into production.
Most evaluation conversations settle around the same topics: isolation, backups, quotas, auth, and whether the platform removes enough operational burden to be worth using instead of homegrown glue.
What teams ask before going live.
Billing is based on API calls, stored memories, and document parses. Neuralbase does not expose token, embedding, or vector-operation billing in the customer surface.
Dashboard access is session-based, runtime access is API-key based, and the vector layer is isolated at the per-user Qdrant collection boundary. Organization membership and key scope still control which workspace can act on data.
No scheduled backups run on the free plan. All plans use persistent storage, but daily snapshots with 30-day retention are a paid feature on Pro and Enterprise only.
Free accounts hard-stop at their plan limits. Pro accounts receive soft quota warnings and can continue past monthly allocation while the team decides whether to upgrade. Enterprise limits are negotiated.
Yes. Document parsing and memory storage are separate actions. You can parse a document, inspect the extracted output, and choose later whether to store it as memory.
The dashboard uses signed server sessions. The runtime API uses scoped API keys. Both paths flow into the same billing and quota layer so usage attaches to the correct account.
Data stays on persistent storage. VM restarts do not wipe memories or document results. Scheduled snapshots are additional protection for paid plans, not the main persistence layer.
Free sign-up is self-serve. Paid activation is currently handled directly while checkout wiring is finalized. If you need Pro or Enterprise access now, use the support link on the pricing page.