Simple, honest pricing.

Credits are the single currency behind every model, so a cheap model and a flagship model can live in the same wallet. Each model has its own published rate per 1,000 tokens (input and output priced separately), shown in the app next to each model. A quick chat with a small model costs a fraction of a credit; hammering the most expensive frontier model costs more. Your balance sits in the status bar and ticks down live as the AI replies, so there is never a surprise invoice. New accounts start with 3,000 credits, and daily check-in adds another 100 for the first 15 days.

They never expire, and unused monthly credits stay in your balance. When your balance hits the floor the app automatically drops back to Free mode — local models and free OpenRouter models keep working, paid models pause, and a prompt offers a top-up. You are never silently charged and nothing is auto-renewed without you turning it on.

Your own key runs in Free mode, on your own quota, forever, at no charge from us. What a membership adds is different: bundled credits at a lower per-credit price, and access to the tiers of models the app otherwise gates — flagship reasoning, media generation and the engineering toolchain. In short, your key pays your provider; a membership unlocks the app.

Once payment completes you'll receive a license key by email. Open the app, switch the mode dropdown to your plan, and paste the key. Activation is bound to one device; reinstalling on the same PC keeps working. Locked tiers show a 🔒 in the dropdown until your subscription or license is active.

Visa, Mastercard and PayPal. Payments are processed securely by Creem, our merchant of record; we never see or store your card details.

No. We don't bolt extra refusal lists or filters onto the models — you get the raw capability you're paying for. Standard provider safety still applies, and you're responsible for how you use the output.

Yes. Your plan stays active until the end of the paid period, then the app simply returns to Free mode. Your chat history is stored locally and never lost.

Completely. Chat history, memories and settings are stored only on your computer. In Free mode nothing ever leaves your machine.

You get 78 built-in models across five plans, spanning DeepSeek, Qwen, GLM, Kimi, MiniMax, Doubao, Hunyuan, Ernie and Grok, plus the OpenRouter catalogue. Some closed models from OpenAI, Anthropic and Google are unavailable in certain billing regions. The app probes this automatically at startup (once per key, at no meaningful cost), badges the affected models as region-locked, and routes around them — so you are never left guessing why a model will not run. In mainland China it goes one step further: Auto mode prefers the domestic model group (DeepSeek / Qwen / GLM / Kimi and friends) so you keep full speed without a VPN. Fourteen China-direct platforms are also preconfigured, covering exactly the models that international aggregators restrict.

Yes — and it is encouraged. Any OpenAI-compatible platform works: SiliconFlow, Alibaba Bailian, DeepSeek official, Volcano Ark, Zhipu, Moonshot Kimi, MiniMax, Tencent Hunyuan, ModelScope, 302.AI, DMXAPI, AIHubMix, NVIDIA NIM, StepFun or OpenRouter. Pick the platform from a dropdown, the base URL fills itself in, paste your key and you are running on your own quota. You can add two platforms at once and the app will fail over between them automatically.

Credits burned by the calls you actually make. Each model is metered at its published credit rate, which is derived from the provider's live price with a small routing margin — there is no monthly message ceiling and no per-seat charge. You start with your 3,000 welcome credits; after that you top up manually or let a linked card auto-charge. Every call logs model, input/output tokens and credits spent, so you can reconcile the balance against your provider bill line by line.