# OmniRoute

> A free, open-source AI gateway that unifies 160+ providers behind one endpoint with smart 4-tier auto-fallback, prompt compression saving 15-75% tokens, and multi-platform support.

OmniRoute is a free, open-source AI gateway that routes requests across 160+ AI providers through a single OpenAI-compatible endpoint, with zero downtime via smart 4-tier auto-fallback (Subscription → API Key → Cheap → Free). It features a built-in 5-mode prompt compression pipeline that saves 15-75% tokens automatically on every request, a 3-level proxy system for bypassing geographic restrictions, and full MCP Server and A2A Protocol support. The project is written in 100% TypeScript, backed by 4,690+ automated tests, and runs on Web, Desktop (Electron), Android (Termux), and as a PWA.

- **Smart 4-Tier Auto-Fallback** — *Install via `npm install -g omniroute`, then configure a combo chain in the dashboard to automatically route Subscription → API Key → Cheap → Free providers with zero downtime.*
- **Prompt Compression (5 Modes)** — *Enable in Dashboard → Settings → Compression; choose from Off, Lite (~15%), Standard/Caveman (~30%), Aggressive (~50%), or Ultra (~75%) to reduce token usage on every request transparently.*
- **160+ Provider Support** — *Connect OAuth providers (Claude Code, Codex, Gemini CLI, GitHub Copilot, Cursor), API key providers (OpenAI, Anthropic, DeepSeek, Groq, xAI, Mistral, and 120+ more), and 11 genuinely free providers including Kiro, Qoder, Pollinations, and LongCat.*
- **Format Translation** — *OmniRoute automatically translates between OpenAI, Claude, Gemini, and Responses API formats so any OpenAI-compatible tool works with any provider.*
- **MCP Server (29 Tools)** — *Run `omniroute --mcp` for stdio transport, or connect via SSE/HTTP; supports 10 granular permission scopes and a SQLite-backed audit trail.*
- **A2A Protocol** — *Full JSON-RPC 2.0 + SSE streaming agent-to-agent protocol with task lifecycle management and an agent card at `/.well-known/agent.json`.*
- **3-Level Proxy System** — *Configure global, per-provider, or per-API-key proxies in the dashboard; includes a built-in 1proxy free marketplace with 500+ community-shared proxies and TLS fingerprint spoofing.*
- **10 Multi-Modal APIs** — *One endpoint covers chat completions, embeddings, image generation, video, music, audio TTS/STT, reranking, moderation, and web search.*
- **Full Observability** — *Unified 4-tab log dashboard (request, proxy, audit, console), p50/p95/p99 telemetry, per-token cost tracking, budget controls, and a built-in LLM evaluation framework.*
- **Multi-Platform** — *Runs as an npm global package, Docker image (AMD64 + ARM64), Electron desktop app, Android via Termux, or as a PWA; available in 40+ languages with RTL support.*

## Features
- Smart 4-tier auto-fallback (Subscription → API Key → Cheap → Free)
- 5-mode prompt compression pipeline (Off / Lite / Standard / Aggressive / Ultra)
- 160+ AI provider integrations
- Format translation (OpenAI ↔ Claude ↔ Gemini ↔ Responses API)
- MCP Server with 29 tools and 3 transports (stdio/SSE/HTTP)
- A2A Protocol (JSON-RPC 2.0 + SSE streaming)
- 3-level proxy system with TLS fingerprint spoofing
- 1proxy free marketplace (500+ community proxies)
- 10 multi-modal APIs (chat, embeddings, images, video, music, TTS, STT, reranking, moderation, web search)
- OAuth PKCE auto token refresh for 8 providers
- Multi-account round-robin per provider
- Circuit breakers with exponential backoff and anti-thundering herd
- Semantic + signature two-tier cache
- Unified observability dashboard with p50/p95/p99 telemetry
- Built-in LLM evaluation framework with golden set
- Cost analytics and budget controls
- 13 routing/balancing strategies
- Cloudflare / Tailscale / ngrok tunnel support
- Desktop app (Electron), Android (Termux), PWA
- 40+ language UI with RTL support
- 4,690+ automated tests across 517 files
- WebSocket bridge (/v1/ws)
- Request deduplication and idempotency
- OpenAPI 3.0 live spec with Try-It UI

## Integrations
OpenAI, Anthropic (Claude), Google Gemini, DeepSeek, Groq, xAI (Grok), Mistral, OpenRouter, GitHub Copilot, Cursor, Cline, Claude Code, Codex CLI, Gemini CLI, OpenClaw, Kilo Code, Windsurf, Continue, LM Studio, Ollama, vLLM, Llamafile, Docker Model Runner, NVIDIA NIM, Cerebras, Cohere, Perplexity, AWS Bedrock, Azure OpenAI, Vertex AI, Cloudflare Workers AI, HuggingFace, Together AI, Fireworks, SambaNova, Kiro AI, Qoder, Pollinations AI, LongCat, ElevenLabs, Deepgram, AssemblyAI, Stability AI, Fal.ai, Runway, ComfyUI, Tailscale, Cloudflare Tunnel, ngrok, Docker, Caddy

## Platforms
WINDOWS, MACOS, LINUX, ANDROID, IOS, WEB, API, VSC_EXTENSION, CLI

## Pricing
Open Source

## Version
v3.7.8

## Links
- Website: https://omniroute.online
- Documentation: https://github.com/diegosouzapw/OmniRoute/tree/main/docs
- Repository: https://github.com/diegosouzapw/OmniRoute
- EveryDev.ai: https://www.everydev.ai/tools/omniroute