OmniRoute — единый AI-шлюз для 1200 моделей

★ 65.5K

OmniRoute is a self-hosted, MIT-licensed AI gateway that exposes a single API endpoint across 352+ providers and 1200+ language models, including Claude, GPT, Gemini, DeepSeek, Kimi, Qwen, GLM, and MiniMax. You reach for it when you're tired of hardcoding separate keys and endpoints for every model, when you need automatic provider fallback on quota exhaustion, when you want to cut token costs without touching application code, or when building a multi-agent system that needs a unified routing layer. Under the hood it features quota-aware auto-fallback, built-in token compression via RTK and Caveman algorithms (saving 15–95% of tokens), MCP and A2A protocol support for agentic pipelines, a Desktop app, and a PWA. Compatible with Claude Code, Codex, Cursor, OpenCode, Cline, and Copilot. Runs as a Desktop app, PWA, or via CLI/API. Built by a community of 550+ contributors. Best suited for teams building LLM-powered products who want to reduce single-vendor lock-in. Overkill if you only target one provider with no redundancy needs. MIT license means it's fully usable in commercial products and resellable as a managed service.