About Ferro Labs
We're building the infrastructure layer that makes AI reliable, observable, and cost-effective at every scale.
Every production AI application needs the same plumbing: load balancing across providers, intelligent retry and fallback logic, semantic caching, cost tracking, and a single audit trail. Today, teams build this themselves — or they don't, and they pay for it with outages and ballooning API bills.
Ferro Labs exists to solve this once, as open infrastructure, so every team — from a solo developer to a Fortune 500 AI team — can focus on what they're actually building.
Principles
Our values
Open by default
The AI Gateway core is open source under the Apache 2.0 licence. Every provider integration, routing strategy, and plugin ships in the public repo — no features locked behind a paywall.
Performance first
Written in Go for minimal overhead. Sub-millisecond routing decisions, concurrent provider fan-out, and a semantic cache that eliminates redundant LLM calls.
Privacy by design
API keys are stored as SHA-256 hashes. Request payloads are never used for training. Compliance certifications (SOC 2 Type II) are on our roadmap, not an afterthought.
Developer experience
One OpenAI-compatible endpoint, zero SDK rewrites. Drop-in compatibility means your existing code works on day one — and you get retry logic, fallbacks, and observability for free.
AI Gateway is open source
Browse the code, report issues, contribute providers and plugins, or self-host on your own infrastructure.