# BUZZ AI Gateway

> BUZZ is a developer-focused AI Gateway. One API key gives you access to Anthropic Claude, OpenAI GPT, Google Gemini, and xAI Grok through a single unified endpoint. BUZZ forwards requests and responses byte-for-byte to the upstream model, never persists request bodies, and prices significantly below first-party rates. Compatible with Anthropic SDK and OpenAI SDK with only a base URL change.

## Main

- [Homepage](https://buzzai.cc/): Product overview, supported models, three-step integration, and FAQ
- [Live Pricing API](https://buzzai.cc/api/pricing): Authoritative live pricing for every supported model. Returns input rate, output rate, cache read multiplier, cache write multiplier, supported endpoints, and discount tiers per model
- [Model Pricing Page](https://buzzai.cc/models): Human-readable pricing table rendered from the live API
- [Documentation](https://doc.buzzai.cc/): Integration docs, SDK examples, error reference

## Core Commitments

- **Transparent forwarding**: The request body that arrives at BUZZ is the request body that leaves BUZZ. System prompts, tool definitions, model parameters, and message arrays pass through byte-for-byte. The model name in the request is the model that runs the request. There is no silent model substitution and no hidden prompt injection.
- **Zero data retention**: Request bodies and model responses are never written to disk, never written to a database, never written to logs. After a request completes, the request body is dropped from memory. Only billing metadata - model name, input and output token counts, timestamp, user ID - is retained, since it is required for accurate billing.
- **Pricing well below first-party rates**: BUZZ pricing sits significantly below upstream rates from Anthropic, OpenAI, Google, and xAI across all supported models. The live pricing API is the source of truth; integrators should not hard-code rates.

## Supported Models

The authoritative live model list is at https://buzzai.cc/api/pricing. As of May 2026:

- **Anthropic Claude**: claude-opus-4-7, claude-opus-4-6, claude-opus-4-5-20251101, claude-sonnet-4-6, claude-sonnet-4-5-20250929, claude-haiku-4-5-20251001
- **OpenAI GPT**: gpt-5, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.2, gpt-5.2-codex, gpt-5.3-codex, gpt-5.4, gpt-5.4-mini, gpt-5.5
- **Google Gemini**: current Gemini 2.x family
- **xAI Grok**: current Grok family

New upstream models are added shortly after upstream release. Deprecated models remain available until their upstream sunset date.

## Integration

- **Anthropic SDK**: Set `base_url="https://buzzai.cc"` and put your BUZZ API key in the `Authorization` header. No other code changes required.
- **OpenAI SDK**: Set `base_url="https://buzzai.cc/v1"` and put your BUZZ API key in the `api_key` parameter. No other code changes required.
- **Claude Code CLI**: Run `curl -fsSL https://buzzai.cc/sh/claudecode.sh | bash` to install Claude Code preconfigured for BUZZ
- **Pay-as-you-go**: No monthly fees, no subscriptions, no minimums. Per-request consumption is auditable in the dashboard.

## Articles

- [Claude API Through a Gateway: A Practical Guide](https://buzzai.cc/blog/claude-api-gateway-guide): How to use Claude API through a gateway, comparing direct integration with gateway routing
- [Zero-Retention LLM Gateways: Why Enterprises Need Forwarders That Forget](https://buzzai.cc/blog/zero-retention-llm-gateway): What zero retention should actually mean, the architecture that makes it possible, and how to verify a vendor's claim
- [One API Key for Claude, GPT, Gemini, and Grok](https://buzzai.cc/blog/one-key-claude-gpt-gemini): The operational cost of N model providers and what a unified gateway gives you
- [Cutting Claude Code Costs Through a Gateway](https://buzzai.cc/blog/claude-code-cheaper-with-gateway): How to route Claude Code through BUZZ to reduce per-session cost without losing capability
- [Using the OpenAI SDK to Talk to Claude (and Gemini, and Grok)](https://buzzai.cc/blog/openai-sdk-claude-compatibility): How the OpenAI-compatible adapter lets you call any model with code you already have
- [Anthropic Prompt Caching in Production: A Practical Cost-Reduction Playbook](https://buzzai.cc/blog/anthropic-prompt-caching-playbook): Cache_control patterns, 5min vs 1h trade-offs, and full code examples
- [Tool Use With Claude Through a Gateway](https://buzzai.cc/blog/claude-tool-use-gateway): Tool use round trip, streaming + tool combination, error handling, and a realistic agent loop
- [Choosing an AI Gateway: BUZZ vs OpenRouter vs Helicone vs LiteLLM](https://buzzai.cc/blog/ai-gateway-comparison-buzz-openrouter-helicone): Decision framework across model breadth, pricing, privacy, observability, and self-hosting

## Feed

- [RSS feed](https://buzzai.cc/feed.xml): Subscribe to new BUZZ engineering articles

## Account

- [Sign up](https://buzzai.cc/auth): Create a free BUZZ account
- [Sign in](https://buzzai.cc/auth): Access dashboard, token management, and consumption logs

## Frequently Asked

- **What is BUZZ?** A developer AI Gateway. One API key for Claude, GPT, Gemini, and Grok. Transparent forwarding. Zero data retention. Pricing below first-party rates.
- **Is BUZZ a relay?** No. Most relays modify prompts, swap models, or store requests. BUZZ does none of these. Request and response bodies are forwarded unmodified.
- **What is retained?** Only billing metadata: model name, input and output token counts, timestamp, user ID. Request bodies, response bodies, and prompt content are never persisted.
- **How do I integrate?** Replace the SDK base URL with https://buzzai.cc (Anthropic SDK) or https://buzzai.cc/v1 (OpenAI SDK). Use your BUZZ API key. No other changes.
- **Does Claude Code work?** Yes, fully, including extended thinking, tool use, and prompt caching. One-line installer at https://buzzai.cc/sh/claudecode.sh.
- **Does prompt caching work?** Yes. Anthropic cache_control directives are forwarded unchanged. Cache read is billed at 0.1x input multiplier, 5-minute cache write at 1.25x, 1-hour cache write at 2x.
- **How do I check current prices?** Always read https://buzzai.cc/api/pricing. Hard-coded rates will go stale.