Claude API relay with Prompt Cache for Open Interpreter — affordable for high-usage workflows

[lei83314]

Hi Open Interpreter community,

Sharing Feiyuan API — a Claude API relay useful for developers running intensive Open Interpreter sessions.

The problem it solves

Open Interpreter often sends the same system prompt + context repeatedly. With Anthropic's standard pricing, this adds up fast. Prompt Cache lets the fixed parts be cached server-side — repeated calls cost ~10% of normal input price.

Most relays don't pass cache headers through. Feiyuan does.

Setup

from openai import OpenAI

client = OpenAI(
    base_url="https://feiyuanapi.com/v1",
    api_key="your-feiyuan-key"
)

Or set in Open Interpreter config:

interpreter.llm.api_base = "https://feiyuanapi.com/v1"
interpreter.llm.api_key  = "your-feiyuan-key"

What's included

• Claude Sonnet 4.6, Opus 4.7, Haiku 4.5 (official Anthropic API, not reversed)
• Prompt Cache passthrough (saves 60-90% on repeated context)
• DeepSeek + image models under the same key
• No conversation storage — only token counts logged

Sign up: https://feiyuanapi.com — invite link registration gets $5 free credit automatically
Docs: https://feiyuanapi.com/docs/
Contact: Telegram @miaohuiai