Billing

Dashboard → Billing loads /api/billing/info, /api/billing/balance, and /api/billing/auto-recharge.

What you pay for

• Pay-as-you-go on the Free plan — token cost from pricing_rules (USD per 1K in the DB, shown per 1M in the UI)
• Pro / Team / Enterprise — $12 / $18 per user / $45 per user. Signup credit is $3. Stripe Checkout uses STRIPE_PRO_PRICE_ID and STRIPE_TEAM_PRICE_ID. Each paid invoice grants a monthly inference stipend ($3 / $5 per seat / $8 per seat) onto prepaid credits. Tokens after that and every GPU-second are pay-as-you-go. Enterprise is talk-to-sales. If those env vars are missing, the button tells you so; it does not pretend checkout worked
• Agents add-on — $20/month (STRIPE_AGENTS_ADDON_PRICE_ID). Separate subscription from the seat plan. Included on Enterprise. Tokens still bill quota
• Web Search add-on — $20/month (STRIPE_WEB_SEARCH_ADDON_PRICE_ID). Live Google results via Vertex AI Grounding on OpenDoor’s GCP project. Customers subscribe; they do not paste GCP keys. Included on Enterprise. Without the add-on, POST /v1/plugins/web-search, workflow Run, and assistant web search return 402 addon_required — they do not fall through to Tavily
• Credit top-ups — Stripe payment, then credit_transactions
• Auto-recharge — when the credit balance crosses your threshold

The marketing Pricing page (/pricing on the dashboard app) reads the same pricing_rules + gpu_skus tables (no login).

Spend controls

Org monthly budget, per-key spend caps, and plan allowances are enforced on the gateway before a provider call. See Spend caps and Auto-recharge.

Assistant monetization (public assistants) uses catalog prices for the AI-cost estimate. If a model has no pricing_rules row, the preview says so instead of inventing GPT-4o rates.