FAQ
Pricing, data handling, latency, residency, languages, and on-prem. If your question isn't here, drop us a line.
Pricing and limits
What's included in Free?
Generation against a capped quota, public-persona-only mode, and basic chat without drift gating. The OpenAI-compatible endpoint is on. The full breakdown lives in tiers.
How is usage metered — tokens, seats, or generations?
It depends on the surface. Generation is metered per persona created and per refine call. Chat is metered per token (input + output). Collaboration is per seat above the per-tier included count. API calls roll into the workspace rate limit.
Is there an annual discount?
Yes. Annual prepay drops the monthly rate by a meaningful percentage on Pro, Team, and Enterprise. See moonborn.co/pricing.
Can I cancel anytime?
Yes. Self-serve in Settings → Billing. Downgrades take effect at the end of the current billing cycle; data is retained per the retention rules of your new tier.
Data and compliance
Is my data used to train models?
No. Customer data — prompts, completions, persona definitions, chat history — is never used to train any model. Customer data is isolated per organization and never crosses tenant boundaries.
Where is data stored? Can I choose a region?
Production data is stored in either the US or EU. The region is selected at signup and locked thereafter; cross-region reads are blocked at the database layer. Enterprise customers receive a written data-residency commitment.
Are you SOC 2 / GDPR / HIPAA?
GDPR — yes; we are the data controller for marketing data and a processor for customer data. SOC 2 — Type II in progress; we will publish the report when complete. HIPAA — available on Enterprise contracts on request.
Performance
What's typical first-token latency?
300–800 ms p50 for streaming chat, varying by upstream model and workspace region. Persona generation is asynchronous (it runs a multi-step pipeline) and surfaces progress via SSE; expect 6–15 seconds end-to-end.
Customization
Can I run Moonborn on-prem?
Self-hosted Moonborn is an Enterprise option. The artifact ships as a containerized stack; you provide compute, storage, and an LLM provider (or use ours via a private link). Contact sales for the dossier.
Languages
Which languages do personas support?
Personas declare an ISO-639-1 language tag and the generation pipeline supports a broad set. English, Turkish, German, Spanish, French, and Portuguese are first-class — drift detection has language-specific baselines for these. Other languages work and are constantly being calibrated.
Still have a question?
Email hello@moonborn.co, or open a discussion on our GitHub.