Mikan Cloud
立即开始

Quickstart

Send your first request to Mikan Cloud in under 60 seconds — curl, Python (OpenAI SDK), or Node.

Three steps: grab a key, point any OpenAI-compatible client at https://api.mikancloud.com/v1, send a request.

  1. 1

    Create an API key

  2. 2

    Sign in, open Dashboard → API Keys, click New key. The full secret is shown once in the format sk-mk-…. Copy it now — we only store an argon2 hash.

  3. 3
  4. 4

    Send a chat request

  5. 5

    The endpoint is OpenAI-shape, so any client that speaks OpenAI Chat Completions works. Set the base URL and you're done.

  6. 6
    curl https://api.mikancloud.com/v1/chat/completions \
  -H "Authorization: Bearer $MIKAN_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v3.2",
    "stream": true,
    "messages": [
      {"role": "user", "content": "Explain SSE in two sentences."}
    ]
  }'
  7. 7

    Send an embeddings request

  8. 8

    The same SDK speaks /v1/embeddings. The cheapest multimodal-capable model is qwen3-vl-embedding-8b (Mikan-hosted, 768d):

  9. 9
    curl https://api.mikancloud.com/v1/embeddings \
  -H "Authorization: Bearer $MIKAN_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-vl-embedding-8b",
    "input": ["alpha", "beta", "gamma"]
  }'
  10. 10

    input accepts string | string[]. Multimodal block-array input (image / video blocks) is on the roadmap — for now, send text only.

  11. 11

    Use it from Cursor

  12. 12

    Open Cursor → Settings → Models. Add a custom OpenAI-compatible model with:

  13. 13
    • Base URL — https://api.mikancloud.com/v1
    • API Key — your sk-mk-… secret
    • Model — deepseek-v3.2 (or any model from GET /v1/models)
  14. 14

    Toggle the new model on. Inline edits, chat, and Cmd-K all flow through Mikan.

What it costs

Each chat request decrements your wallet by (prompt_tokens × in_rate + completion_tokens × out_rate). Each embedding request decrements by total_tokens × rate. The 7-day token & cost graph on your dashboard updates in near-real-time.

next →API Reference