Hey Product Hunt 👋 I started OmniRoute because every AI coding tool, provider, account, quota, and wire format felt like a separate integration. A rate limit, exhausted quota, or unavailable model could stop a coding session even when another healthy route was available. OmniRoute turns that fragmented setup into one self-hosted endpoint you control. Point Claude Code, Codex, Cursor, Cline, Copilot, or another compatible client at it, then configure the providers you want. OmniRoute handles quota-aware routing and fallback at the provider, connection, and model levels. What makes it different: • Free and MIT-licensed • Self-hosted and local-first • OpenAI, Anthropic, Gemini, and Responses API compatibility • Automatic fallback and cost/quota-aware routing • 12-engine token compression pipeline • Local dashboard for providers, health, usage, and costs • Built-in MCP and A2A support The project began as a practical router for coding agents and evolved through real failure cases and contributions from an open-source community into a broader AI gateway and control plane. OmniRoute is now part of Cheaper Inference, while the self-hosted project remains free and MIT-licensed. Quick start: npm install -g omniroute I’d especially value candid feedback: 1. Is the one-endpoint + automatic-fallback value clear? 2. Which setup or integration still feels too difficult? 3. What would you need to trust OmniRoute for everyday use? Thanks for trying it—and please tell us what breaks or feels confusing.

OmniRoute

Self-hosted AI gateway with automatic provider fallback

Self-hosted AI gateway with automatic provider fallback