Billing & Plans

Current plan
Developer Free forever
164 / 500 credits used this cycle
Credits & alerts
Auto top-up when credits run low · purchases automatically below 50 credits
Plan comparison — subscriptions unlock capabilities and bundled quotas
Developer Current planPro RecommendedScale Enterprise
Free forever$49 / mo$259 / moCustom pricing
10 end users · never expires1,000 end users10,000 end usersTiered MAU pricing · committed-use discounts
Unlimited projects · unlimited agentsUnlimited projects · unlimited agentsUnlimited projects · unlimited agentsUnlimited projects · unlimited agents
500 credits / mo10,000 credits / mo100,000 credits / moSLA · dedicated support
Standard models · text + TTSStandard + Pro models · realtime voiceStandard + Pro models · realtime voiceCustom model integrations · private deployment
Community supportTicket support · webhooksPriority support · webhooksInvoicing & corporate settlement
CurrentContact us
Per-end-user fair use — stops a few users from burning through your quota: every end user gets a monthly base allowance, and overages follow the policy you choose
MAU definition: an external_user_id only counts if it had ≥1 session that month — inactive users are never billed; session-level rate limits prevent single-user abuse.
Usage pricing — all consumption converts to credits (numbers are illustrative)
Billable itemUnitCreditsUnderlying cost
Chat · standard modelsper 1K tokens1 creditLLM calls
Chat · Pro modelsper 1K tokens5 creditsLLM calls
Speech synthesis (TTS)per 1K characters2 creditsAudio
Realtime voice callsper minute10 creditsAudio
Additional end users (beyond plan MAU)per user · month20 creditsMemory storage + embeddings + retrieval
Knowledge-base storageper GB · month50 creditsMemory storage
Why we bill this way — how the three cost drivers price in
1 End-user count (MAU) is the only hard gate Projects and agents are just configuration with near-zero cost, so they're unlimited; memory (storage, embeddings, retrieval) scales with users, so it's bundled into MAU — user count is cost count, and it's the easiest number for customers to forecast.
2 LLM calls → per token The largest variable cost, with different credit multipliers per model tier; developers can pick standard or Pro models per agent to control spend.
3 Audio → per character / minute Speech synthesis bills per thousand characters, realtime calls per minute. Everything else (A2A, instant generation) costs little and is absorbed into the base price rather than billed separately.
Invoices
PeriodAmountStatus
2026 年 6 月$1.69Paid
2026 年 5 月$0.00Paid

UI preview — plans, unit prices and invoices are illustrative; sign in for your real billing status.