Zaguán
Sign InGet Started
Overview
Version 0.27.0 - November 2025

One OpenAI-Compatible API. Every Major LLM Provider.

Zaguán is an OpenAI-compatible gateway with a smart translation layer built for solo devs, small teams, and agencies. Point your SDK at a single endpoint, swap any `provider/model` identifier, and ship in minutes - our translation layer handles the complexity of adapting your requests to each provider's native format.

November 2025 Release: 9 major features including OpenAI Realtime API, audio output, xAI stateful conversations, Anthropic prompt caching (90% cost savings), Google Gemini thinking, and our smart extra_body translation layer that lets you access any provider-specific feature through OpenAI-compatible syntax.

November 2025 Release Highlights

🎭

Realtime Voice API

WebSocket-based voice conversations with ultra-low latency

🔊

Audio Output

Generate spoken audio directly from chat completions

🔗

xAI Stateful Conversations

30-day server-side storage with encrypted thinking

💡

Prompt Caching

90% cost savings with Anthropic Claude caching

🧑‍💻

Gemini Thinking

See the model's reasoning process in real-time

🎯

Extra Body Translation

Access any provider-specific feature seamlessly

🌐

Virtual Models

Multi-provider routing with automatic failover and load balancing

What you get

  • •
    OpenAI-compatible API: Use familiar OpenAI endpoints and request schemas. Our translation layer adapts your requests to each provider's native format, so swapping openai/gpt-4o-mini foranthropic/claude-3.5-sonnet just works.
  • •
    Smart translation layer: We automatically adapt prompts, parameters, and formats to each provider's API - you don't need to learn every vendor's quirks.
  • •
    Managed infrastructure: Hosted routing layer with observability, authentication, and traffic controls built-in.

What we do not do

  • •
    No hidden model tuning: Zaguán forwards your payloads as-is; custom prompting strategies stay under your control.
  • •
    Transparent error handling: Standard models expose consistent errors so you control retry logic. Virtual Models provide automatic failover when you need it.
  • •
    No proprietary lock-in: All endpoints follow documented standards so you can migrate at any time.

Platform highlights

  • •
    Translation layer: We don't just proxy requests - we actively adapt OpenAI-style calls to work optimally with each provider's specific API, parameters, and features.
  • •
    Security posture: Incoming requests are authenticated with Zaguán API keys and proxied through dedicated provider integrations - your provider credentials never leave our environment.
  • •
    Built for small teams: Simple, predictable pricing at 15 € / month for early adopters (Founder's Plan locked in forever), 39 € / month standard pricing. No enterprise contracts required to start.

Next steps

  1. Review the API guide to point your OpenAI SDKs at https://api.zaguanai.com/v1.
  2. Browse available Models and choose the namespaced identifier that fits your workload.
  3. Explore Virtual Models for multi-provider routing with automatic failover and load balancing.
  4. Request early access through the Founder's Plan CTA on the homepage to receive your API key.