Tiers — what runs where
Loading…
How a message is routed
first match wins · deterministic · zero LLM cost
Try a message
dry decision — nothing is sent to a model
Routing settings
Model thinking (reasoning) per model
thinking tokens are billed as output and count against max tokens — always set explicitly
modelthinking in the brain loopprovider pin
Background calls (auto-facts, classifier) always run with thinking off/minimal. Saved with “Save routing”.
Recent decisions
last turns · tier · reason · model actually used · tokens · cost
timemessagetierwhymodelcallstokens in / outcostms
Loading…