All Tools
S
LLMFreeOpen Source
SWITCHYARD
Route LLM traffic across models without changing client APIs
Apache-2.0
ABOUT
Sending every agent turn to the strongest model wastes money, while a cheap default fails hard tasks. Switchyard inspects the request or recent tool activity and picks a model from your pool so gateways can mix efficient and capable models without rewriting clients.
INTEGRATION GUIDE
1. Route coding-agent turns to a cheap model and escalate when tools fail
2. Plug Switchyard into LiteLLM or NeMo Relay with OpenAI-compatible clients
3. Benchmark Auto vs Execution routing against a single-model cost baseline
4. Embed Switchyard routing in a custom Rust or Python agent harness
TAGS
llmroutingnvidianemoopenai-compatibleanthropicgateway