Start free with 1,000 credits a month. Move up when your agents are in production. Switch models any time without losing what they learned.
Operations have documented credit weights. Managed-model calls additionally report token and estimated-cost usage. Your dashboard exposes the quota impact, while stored memory remains portable across models.
Credits are normalized usage units. Operations have documented credit weights, while managed model calls also track token and model usage. Your dashboard shows both quota and usage.
On Free and Starter, memory writes pause until the next cycle and reads keep working. On Pro and Scale you can turn on overage billing, so you are never interrupted and only pay for what you use.
Yes, on Starter and above. Save an Anthropic or OpenAI API key in your dashboard or with POST /v1/byok/providers/{provider}. Once the key validates, every model call Hebbrix makes for your account (chat and answers, reasoning search, memory extraction, profile enrichment, conflict resolution, document and follow-up processing) runs on your key and is billed by your provider. Hebbrix then charges only the platform part of each operation: chat 1 credit for any model, reasoning search 2, smart ingestion 1; all other operations are unchanged. If your provider rejects the key, your provider account is out of credit, or it rate-limits the key, the request fails with an error that names your key; Hebbrix never falls back to its own key. Embeddings and reranking always run on Hebbrix infrastructure. Google keys can be saved but are not used for routing yet. Memory stays model-agnostic and portable.
Always. Export every memory and the knowledge graph at any time, on any plan. Your memory is yours: portable, auditable, and deletable.
Loaded from the same public plan catalog used by Dashboard Billing. Contract-specific Enterprise terms are separate. Graph reads can be available on lower tiers; manual graph mutation and inference have additional gates described in the graph guide.