Overview

Mikan Cloud is an OpenAI-compatible API gateway with a single Stripe wallet, multiple upstream providers, and a self-hosted multimodal embedding model.

One Bearer key reaches OpenAI, Anthropic, Google, DeepSeek, and Mikan's own GPU-hosted Qwen3-VL embedding model. One wallet pays them all in USD, top-up style — no per-vendor accounts, no monthly subscriptions, no minimums.

What's on the gateway

• Chat completions — gpt-4o, gpt-4o-mini, claude-opus-4-7, claude-sonnet-4-6, claude-haiku-4-5, gemini-3-pro, gemini-3-flash, deepseek-v3.2. SSE streaming on every model.
• Embeddings — qwen3-vl-embedding-8b (self-hosted, 768d, multimodal-capable), text-embedding-3-small, text-embedding-3-large.
• OpenAI-compat models endpoint — GET /v1/models returns the live list with capability hints.

If you need a model not on this list, email support — adding a route is a one-table change once we have the upstream key.

Pricing model

Pay-as-you-go from your wallet. Headline frontier models (OpenAI / Anthropic / Google) ship at 0% markup — Mikan only adds a markup on long-tail upstreams (DeepSeek today). Embedding rates are flat per-1M-input-token; chat rates are split into input / output.

See pricing for the live rate card.

Why one wallet

• Cursor users plug Mikan in as a custom OpenAI base URL and get DeepSeek pricing without DeepSeek-only billing or compliance.
• atomos and similar PKM clients call Mikan for both chat and embeddings without juggling two vendor accounts.
• Top-up balances are non-expiring; unused balance is refundable per the refund policy.

Compatibility

OpenAI Python and Node SDKs work unmodified — set base_url / baseURL to https://api.mikancloud.com/v1. Cursor, OpenWebUI, LibreChat, and any LangChain / LlamaIndex OpenAI provider drop in the same way.