Switchyard

Keep AI features running when a model provider has an outage.

An LLM gateway you point your existing OpenAI-style client at. You keep your own provider keys; Switchyard adds model aliases with automatic fallback, per-key rate limits, usage numbers and request logs that leave prompt text out by default.

Is Switchyard a fit?

Pick Switchyard if

  • You call more than one model provider, or want a backup
  • You want no markup on tokens: you pay providers directly

Look elsewhere if

  • You want one bill for model usage (Switchyard never resells tokens)
  • You use a single provider and never need fallback

Use it instead of hand-written retry and fallback code inside every service.

What you get

  • OpenAI-compatible endpoint, including streaming
  • Aliases that fall back to a second provider or model on errors
  • Per-key rate limits that return standard 429 errors
  • Usage and logs by model and provider, without prompt content by default

Honest limits

  • Bring your own provider keys; Switchyard does not sell model access.
  • Free plan keeps 7 days of request logs.
switchyard.kapturisemedia.com
Switchyard app screen from its demo recording
From Switchyard's demo recording, with example data.

Works well with

Part of b2b saas team.

Prices, limits and features on switchyard.kapturisemedia.com are authoritative; this page is a summary checked on 2026-11-17. Prices exclude VAT. UAE customers pay 5% VAT. All prices are in US dollars.