Want your engineers to embed AI Q&A into your product? This covers common integration scenarios, security, and billing — enough for decision-makers to scope a project. Technical details live in the API Reference.
Key takeaways
Common integrations our customers ship: (1) Website chat — visitors ask, AI answers from product docs, cutting human escalation by 60%. (2) Internal knowledge portal — employees query policies, procedures, SOPs. (3) IM bots (Slack / Lark / Teams) — @ the bot for any doc-related question. (4) Sales assistant — reps look up contract terms or pricing during live calls.
API keys support three layers of protection: (a) Bind to a specific KB — the key can only query that KB, can't touch sensitive data; (b) Rate limit — cap per-minute calls to prevent integration bugs from blowing up the bill; (c) IP allowlist — only your specified server IPs can use this key. With all three on, a leaked key is very hard to abuse.
API is billed on actual usage — each call consumes credits = input tokens × rate + output tokens × rate. We publish transparent rate tables per model (see Pricing page); pick a model that fits your budget. When monthly usage exceeds your plan quota, top up pay-as-you-go. Low-balance auto-email alerts prevent service interruption.
Recommended flow: (1) Sign up free, validate your use case in the Playground first — confirm before committing. (2) Pick a plan matching your expected monthly call volume. (3) Owner generates an API key scoped to the production KB. (4) Hand the key to engineers for app integration. Every step is self-serve — unless you need private deployment or custom workflows.