Experiential Labs is an open-source AI gateway whose goal is to unify access to all AI models with a single API key, including hosted providers, users' own keys, and their own GPUs. Users only need to point their agents to the api.experientiallabs.ai/v1 endpoint to call various models from a unified interface, with streaming responses passed through token by token. The gateway itself is free and open source, and users pay the provider's prices with no additional markup; the company monetizes through hosted inference and a Pro subscription.
Main capabilities:
- Unified access and routing: one endpoint aggregates hosted providers, users' own keys, and their own GPU resources, supporting routing, access control, and call attribution.
- Intelligent optimization: observes traffic and suggests when switching models would be better, including adopting new models on the day they are released; provides prompt-specific optimization options; monitors opportunities to improve cache hit rates, and once enabled, repeated tokens enjoy a 90% discount.
- Own models: fine-tuned based on user traffic, validated before launch, accessed through the same endpoint, manually enabled by the user, and not executed automatically. The company says its performance can reach one-tenth the price of frontier models at several times the speed.
- Unified console: view catalog, usage, and limits by agent, person, model, or date; expandable rows show routing, fallbacks, and the actual charges by the provider.
- Team management: one API and one login entry, with permissions scoped by team or key; manage keys, zero-day launch of new models, and place your own fine-tuned and self-hosted models behind the same key.
- Cost control: assign separate keys to each person, each agent, and each tool; administrators set daily, weekly, or monthly limits and enforce them on every request; supports roles, model allowlists, and local-only scopes.
How to use: users change the base URL to api.experientiallabs.ai/v1 and keep their existing code; hand the prompt to Claude Code, Cursor, or Codex to complete setup automatically. They can also choose to self-host the gateway or use the hosted gateway.
Use cases: teams or enterprises that need unified management of access to multiple AI models, control of team AI spending, optimization of model call costs and cache efficiency, and that want to have dedicated models trained on their own workflows.
Limitations: the materials do not provide a specific list of supported providers or models, detailed pricing for the Pro subscription, complete steps for self-hosted deployment, or the specific billing differences between the free gateway and hosted inference.
