Why UniGateway?

UniGateway is a lightweight, high-performance LLM gateway written in Rust. Designed for developers who use multiple AI tools (Cursor, Zed, Claude Code, OpenClaw) and multiple providers (OpenAI, Anthropic, DeepSeek, Groq). One local endpoint, all your models.

Unified, OpenAI-compatible API

Every provider speaks the same protocol. Anthropic, DeepSeek, Groq, local models — they all appear as a single OpenAI-compatible endpoint at localhost:3210. Your tools don't need to know which provider is behind the curtain.

Mode-based routing

Group providers into semantic modes — fast, strong, backup. Switch the active mode with a single command. Each mode can have its own model, provider, and fallback chain.

Failover & stability

Built-in fallback strategies. If a provider goes down, UniGateway automatically routes to the next one. Your AI tools keep working. Zero downtime, zero manual intervention.

Deep visibility

Understand exactly what's happening. ug route explain shows how requests are routed. ug doctor runs a full diagnostic. ug test sends a smoke request. No black boxes.

Pre-configured integrations

Get ready-to-use configuration snippets for your favorite tools. No manual JSON editing, no guessing at field names.

MCP server

Built-in Model Context Protocol server. AI assistants in Cursor or Claude Desktop can manage your gateway through natural language. Run ug mcp and point your tool at it.

AI agent skills

Ships with a Skill file and OpenAPI spec. AI coding agents can read these to understand and automate your entire LLM infrastructure — create services, bind providers, manage API keys, configure routing — all through natural language.

Single binary, zero dependencies

Written in Rust. Compiles to a single static binary. No runtime, no containers, no background services. Install via curl, brew, or cargo and run immediately. Fast startup, minimal memory footprint.