About Experiential Labs
Experiential Labs is an open-source AI gateway that routes requests to over 1,000 models through a single API key. Developers can use the built-in marketplace, bring their own provider keys, or connect models running locally or in their own cloud. The platform analyzes traffic traces to identify wasted tokens, recommend better models, and train specialized models that the user owns.
Review
Experiential Labs enters the AI infrastructure space as a zero-markup gateway with an unusual twist: it learns from your traffic. The tool launched publicly a week ago and has already seen adoption from over 1,000 developers and 50 companies processing more than 10 billion tokens daily. It's built by two AI researchers who wanted a single interface for model access without paying a per-token tax.
Key Features
- Unified API endpoint covering 1,000+ models across marketplace, BYOK, and self-hosted/local setups
- Zero token markup on all requests, with the codebase available as open source
- Traffic analysis that catches cache misses, wasted tokens, and identifies work suitable for async or batch processing
- Training capability that produces specialized models you own, derived from your usage patterns
- Embedded guardrails in the gateway for stripping PII and sensitive content from prompts and responses
Pricing and Value
The gateway itself carries zero markup on token usage. For the launch week, several models are free: GPT-6 Astra, DeepSeek V4 Flash, Qwen 3.8 27B, GPT-5.6 Luna, and Fable 5.1. Payment processing goes through Stripe. The open-source repository is available on GitHub at no cost. Long-term pricing beyond the zero-markup gateway model is not yet defined.
Pros
- No additional cost per token-you pay only what the upstream providers charge
- Single key eliminates managing multiple provider accounts and tracking spend across separate dashboards
- Open-source codebase allows self-hosting and inspection of the routing logic
- Traffic-aware routing can prioritize cache warmth or cost depending on your preference
- Native integrations exist for coding agents like Codex and Claude Code
Cons
- Platform is brand new and lacks a track record for reliability under sustained enterprise workloads
- PII-stripping guardrails are currently in the open-source repo but not yet incorporated into the hosted web platform
- Not well suited for teams that require SOC 2 compliance or contractual data processing agreements out of the box-documentation on these areas is sparse at this stage
Experiential Labs fits small to mid-sized development teams that already juggle multiple model providers and want to consolidate without paying a middleware tax. The traffic-learning component may appeal to teams running high-volume inference who suspect they're burning tokens on cache misses or suboptimal model choices. Organizations with strict compliance requirements or a low tolerance for early-stage tooling should monitor the project's maturity before adopting it in production pipelines.
Open 'Experiential Labs' Website
Your membership also unlocks:
