Experiential

Experiential is an open source gateway and router for AI agent workflows. It gives you one OpenAI-compatible API in front of hosted models, your own provider keys (BYOK), local models, and custom fine-tuned models, so agents and apps call a single endpoint instead of juggling provider SDKs. A setup wizard (pip install experiential, then exp) walks you through connecting providers, issuing keys, and setting a default spend budget, and a hosted version at platform.experientiallabs.ai serves the same API — OpenAI-compatible plus Anthropic Messages — if you don't want to run the gateway yourself.

Beyond proxying requests, Experiential acts as a control plane: you decide which users and agents can use which models, for which use cases, and how much they can spend. It collects OpenTelemetry traces from your agent traffic, and exp build uses those traces to simulate your workload and fit a custom router optimized for quality, speed, and cost. You can go a step further with exp optimize model, which fine-tunes an open source model on your traces via Thinking Machines' Tinker.

Compared to hosted routers like OpenRouter, Experiential is Apache-2.0 licensed and can run entirely on your own machine, so keys and prompts never have to leave your infrastructure. Compared to gateway platforms like Portkey or Cloudflare AI Gateway, it focuses on agent workflows — per-identity budgets, trace-driven routing, and turning production traffic into custom routers or fine-tuned models you own.