Experiential is an open source gateway and router for agent workflows that unifies access to hosted, bring-your-own-key, and local language models through a single OpenAI-compatible API.
The tool addresses the complexity of managing multiple model providers and controlling access across different users, agents, and use cases. It works by accepting traffic through an OpenAI-compatible interface, routing requests to various inference providers, and collecting telemetry from production usage. This traffic data becomes the foundation for optimization: the tool can analyze patterns to recommend better model choices, reduce costs, and train a specialized model tailored to your specific workloads.
Experiential suits teams running agent workflows who want centralized control over model access, spending limits, and provider management without being locked into a single vendor. The local gateway option appeals to those preferring self-hosted infrastructure, while the hosted platform at platform.experientiallabs.ai serves teams wanting managed operation. The tool integrates with existing coding agents like Claude Code, Cursor, and Aider by presenting itself as a drop-in replacement endpoint. Unlike a simple load balancer, it adds identity-based access control, per-user budget enforcement, and the ability to transform production traffic into training data for model optimization.
The project maintains active development with regular updates to its core gateway functionality and API compatibility layers. Work spans both the local gateway implementation and the hosted platform infrastructure, indicating sustained effort across deployment models. The codebase includes tooling for trace collection and model fine-tuning workflows, suggesting ongoing expansion of the optimization pipeline beyond basic routing.
