MixLLMProxy

One endpoint for all your LLM providers. Route by alias, track everything, cap costs.

An OpenAI-compatible proxy that lets you switch between Anthropic, Groq, OpenAI, DeepSeek, and any other provider without touching your application code. Set per-provider daily rate limits to cap spend. Every request is logged to PostgreSQL with a built-in inspection UI.

Under 1000 lines of Haskell. Source on GitHub

What It Does

• Vendor-agnostic routing — Create an alias like gpt-4o → OpenAI or claude → Anthropic. Your app calls MixLLMProxy with the alias name instead of hardcoding endpoints and keys. Swap providers by changing the alias in the UI.
• Cost control — Set daily request or token caps per alias. Requests over budget are rejected automatically. See rolling 24h usage at a glance.
• Full observability — Every request, response, latency, and token count is logged to PostgreSQL. Browse, search, and inspect from the built-in web dashboard.
• Zero client changes — OpenAI-compatible API. Any existing SDK or tool that speaks chat/completions works out of the box.