AI Gateways
Unified endpoints, routing, and fallback across multiple AI model providers.
AI Gateways
AI gateways put one API in front of many model providers, handling authentication, quotas, retries, and fallback. They are useful when you want to switch models without rewriting application code.
This topic covers self-hostable gateways and companion authentication plugins that expose multiple commercial and open-weight models through a single endpoint.
5 projects
- cheaper-inference/OmniRoute — Self-hosted AI gateway that routes one endpoint across 375 providers — Multi-provider routing: Yes; Fallback and retries: Yes; Rate limits and budgets: Yes; OpenAI-compatible endpoint: Yes; Logging and observability: Partial
- diegosouzapw/OmniRoute — Never stop coding. — Multi-provider routing: Yes; Fallback and retries: Yes; Rate limits and budgets: Yes; OpenAI-compatible endpoint: Yes; Logging and observability: Yes; Response caching: Yes; Guardrails and content filtering: Yes
- BerriAI/litellm — Self-hosted AI gateway and Python SDK for 100+ LLM providers — Multi-provider routing: Yes; Fallback and retries: Yes; Rate limits and budgets: Yes; OpenAI-compatible endpoint: Yes; Logging and observability: Yes; Response caching: Yes; Guardrails and content filtering: Yes
- Alph4d0g/opencode-omniroute-auth — opencode authentication provider to authenticate against omniroute — Multi-provider routing: Yes; Fallback and retries: Yes; Rate limits and budgets: Yes; OpenAI-compatible endpoint: Yes; Logging and observability: Yes; Response caching: No; Guardrails and content filtering: No
- Portkey-AI/gateway — Lightweight AI gateway with routing, guardrails, and 1600+ models — Multi-provider routing: Yes; Fallback and retries: Yes; OpenAI-compatible endpoint: Yes; Logging and observability: Yes; Response caching: Yes; Guardrails and content filtering: Yes