Claude API relay with Prompt Cache for Jan — save 60-90% on repeated context

[lei83314]

Hi Jan community,

Sharing Feiyuan API — a Claude API relay that works as a drop-in backend for Jan when you want to use hosted Claude models.

Setup in Jan

In Jan's model settings, add a custom endpoint:

API Base: https://feiyuanapi.com/v1
API Key:  your-feiyuan-key
Model:    claude-sonnet-4-6

Key feature: Prompt Cache passthrough

Jan's system prompts and conversation context get sent on every call. With Prompt Cache enabled:

• First call: normal price
• Subsequent calls with same context: ~10% of input price

For heavy Jan users, this typically cuts Claude costs by 60-90%.

What's available

• Claude Sonnet 4.6, Opus 4.7, Haiku 4.5
• DeepSeek models (cheap reasoning/coding tasks)
• Image generation (gpt-image-2)
• Based on real Anthropic API — not chat2api or reversed tokens

Sign up: https://feiyuanapi.com/?utm_source=github&utm_medium=discussion&utm_campaign=feiyuan&utm_content=open-interpreter — invite link registration gets $5 free credit
Docs: https://feiyuanapi.com/docs/?utm_source=github&utm_medium=discussion&utm_campaign=feiyuan&utm_content=open-interpreter
Questions: Telegram group: https://t.me/feiyuanapi_group