Experiential Labs has released an open source model gateway on GitHub that provides a single control plane for closed, open-source, local, and custom models. Built in Rust for concurrency, it standardizes streaming formats, tool calls, model parameters, rate limits, and error behavior across providers.

The gateway supports every major inference provider and 1000+ models refreshed daily via a codex agent that opens pull requests. It adds under 1 ms for BYOK requests and under 2 ms when Experiential supplies the provider key. Users can deploy it on their own infrastructure or use the hosted version at platform.experientiallabs.ai with zero markup.

Unlike competitors, Experiential is open source, takes no markup, allows mixing local models with a marketplace, and can train a custom model from opt-in production traffic. Routing decisions use standardized OTel traces, simulated rollouts via text world models, an LLM judge, and a nearest-neighbor classifier on prompt embeddings. This approach can deliver a better cost/quality Pareto curve than single-model calling, while also enabling cache hit optimization and new model suggestions.

The hosted platform serves both an Open AI-compatible and Anthropic Messages API at https://api.experientiallabs.ai/v1, with onboarding prompts supporting Claude Code, Cursor, Codex, Aider, and similar coding agents.