Experiential Labs

Experiential Labs is an open source AI gateway that gives you one API endpoint to access dozens of large language models from providers like OpenAI, Anthropic, Google, Meta, DeepSeek, Mistral, and more. You bring your own API keys from those providers (or use Experiential Labs' hosted inference), and the gateway charges zero markup — you pay exactly what the provider charges. The same endpoint also supports models you host on your own GPUs or via services like Bedrock and Azure AI Foundry.

What makes Experiential Labs useful is the intelligence layer sitting on top of the routing. It monitors your traffic and suggests cheaper or faster models that perform equally well for specific prompts. It spots caching opportunities — repeated tokens come back at 90% off when you enable it. And you can fine-tune a small model on your own workflow, prove its quality in a closed-loop simulation, then serve it through the same API. The console gives you per-agent spend, per-model latency, cache hit rates, and real-time request logs. You can set caps per key by day, week, or month, and define model allowlists so an intern's key only hits local models.

This tool is built for developers and AI teams who manage multiple model subscriptions and want to centralize access, control costs, and optimize performance without vendor lock-in. Typical use cases include running a support agent that uses a cheap flash model for simple queries and falls back to a frontier model for complex ones, or fine-tuning a 4B parameter model for claim research that matches a large model's accuracy at a tenth of the price. If you already use OpenAI’s or Anthropic’s SDKs, you just change the base URL — the rest of your code stays the same.