Getting Started

AI Setu is an OpenAI-compatible AI gateway. One workspace API key authenticates calls to many upstream LLM providers, with model-based routing, cross-provider failover, bring-your-own-key (BYOK), response caching, and credit accounting on top.

Point any OpenAI-compatible SDK — or the drop-in @ai-setu/client — at https://gateway.aisetu.ai/v1, change nothing else, and your existing code works against every provider AI Setu supports.

Where to start

• Quickstart — you're a person integrating AI Setu into an app or script. Get an API key and make your first call.
• Agent Quickstart — you're wiring up an autonomous agent that calls AI Setu in production with no human in the loop. Covers ephemeral tokens and spend caps.

Core concepts

Once your first call works, these pages cover how AI Setu makes routing decisions and what's available on top of a plain chat completion:

• Models and Providers — what you can call and how a model id resolves to an upstream provider.
• Routing — resolution order, failover, and optional load balancing.
• BYOK — route through your own provider credentials.
• Caching — exact-match and semantic response caching.
• Guardrails — input/output safety checks.
• Prompts — versioned, server-managed prompt templates.
• Budgets — opt-in spend and token caps.

Reference

• Rate limits, Errors, Data handling — operational details for running AI Setu in production.
• SDKs, MCP, Frameworks — client libraries and integration patterns.