Dictionary term
AI Gateway
Also called LLM gateway · LLM proxy · model gateway
A proxy layer between your applications and one or more model providers that gives a single API and centralises routing, fallbacks, authentication, rate and spend limits, caching and logging for model (and sometimes MCP tool) traffic.
In plain English
One central doorway all your apps use to reach AI models, so you can switch providers, set budgets and see usage in one place.
Analogy
Like a company switchboard: every outside call goes through it, so you can route, limit and log calls without rewiring each phone.
Why it matters
Once several automations call several models, a gateway stops API keys and model choices from being scattered across workflows and gives you failover, per-client spend caps and one usage log.
In practice
Routing every n8n and app call through a gateway so classification goes to a cheap model, drafting to a stronger one, and traffic fails over to a second provider if the first returns errors.
Related terms
Sources
Browse all terms →Last verified