Recipes

Put a calibrated gate in front of any AI action.

Every recipe is the same three moves — score your confidence, gate the action (act / escalate / abstain), and report what actually happened so the cell sharpens. Each is a few lines over the zero-dependency SDK.

/ Auto-merge

Gate auto-merge: let an AI agent ship code only when it's safe

Is this agent's PR safe to ship unsupervised? CI is the verifier — it auto-reports.

Read the recipe→
/ RAG grounding

Stop RAG hallucinations: gate the answer on whether retrieval is grounded

Is the retrieval grounded enough to answer — or should it abstain instead of guessing?

Read the recipe→
/ Tool call

Gate agent tool calls: execute the safe ones, escalate the risky ones

Is this action safe to execute? Scale stakes to reversibility; abstain on irreversible guesses.

Read the recipe→
/ Support reply

Gate support-bot replies: auto-resolve the safe tickets, route the rest

Auto-resolve the ticket, or route to a human? Reopen/CSAT is the verifier.

Read the recipe→
/ Model routing

Calibrated LLM model routing: use the cheap model only when it's trustworthy

Is the cheap model trustworthy enough here, or escalate to the expensive one?

Read the recipe→

Pick a recipe, wrap your first call in about five minutes. Free to start — no card.

Start free →