BUZZ AI Gateway Documentation
Drop-in compatible with the Anthropic, OpenAI, and Google Gemini SDKs. Transparent byte-for-byte forwarding, native Prompt Caching, full Tool Use fidelity, zero data retention.
One call
Three lines and you are talking to Claude.
curl https://buzzai.cc/v1/messages \
-H "Authorization: Bearer $BUZZ_API_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-4-6",
"max_tokens": 80,
"messages": [{"role": "user", "content": "Hello"}]
}'API Reference
Complete reference for every endpoint, with verified request/response shapes.
messages
Anthropic-native chat completion. Streaming, tool use, prompt caching, extended thinking.
POSTchat/completions
OpenAI-compatible alternative. Same Claude / GPT / Gemini access through one schema.
GETmodels
List all models available to your key. Auto-formatted as OpenAI / Anthropic / Gemini.
POSTrerank
Cohere / Jina-style document reranker.
GETuser/self
Profile and quota for the authenticated user.
GETbilling
Subscription quota and recent usage in OpenAI-compatible shape.
GETtokens
Manage your API keys: create, list, rotate, revoke.
Guides
Step-by-step walkthroughs for every common integration path.
Quickstart
From sign-up to your first response in five minutes.
Authentication
Three accepted header forms. Best practices for keys, rotation, allow-listing.
Claude Code
One environment variable to point Claude Code at BUZZ.
OpenAI SDK
Use the official OpenAI SDK to call Claude, Gemini, and Grok.
Streaming
Server-Sent Events with the seven Anthropic event types.
Error Handling
Backoff templates, 529 fallback, error envelope shapes.
Migrate from Anthropic
One-line base_url change. Side-by-side before/after for Python, Node, curl.
Concepts
The mental models behind BUZZ.
Transparent Forwarding
Why byte-for-byte forwarding matters and how BUZZ implements it.
Zero Retention
What is stored, what is not, and why.
Multi-Vendor Routing
Channels, groups, priority and weight, automatic failover.
Prompt Caching
How cache_control works and how to keep your hit rate high.
Tool Use
Function calling shape, tool_choice modes, the multi-turn loop.
Recipes
Production-ready patterns you can copy-paste.