summarise
Long text in, faithful short text out. Length and register are inputs, not code.
inputs: document, max_words
input · llm · output
Eight working flows. Clone one, edit it into yours.
Every template is a real, runnable workflow — declared inputs, real node sequence, publishable as-is. The lane under each shows its nodes in real proportion.
ticket-classifier
Support tickets labelled and routed with a confidence score. The flow used across these pages.
inputs: ticket
input · llm · transform · output
extract-structured
Free text to a typed JSON object that matches your declared schema, or a clean failure.
inputs: text, schema
input · llm · transform · output
meeting-actions
A transcript in, owners and action items out — each with the line that justifies it.
inputs: transcript
input · llm · transform · output
grounded-qa
Answers drawn only from your documents, with the passages used. Says "not found" rather than guessing.
inputs: question, corpus
input · retrieve · llm · output
research-synthesis
Several sources compared into one brief: agreements, contradictions, gaps.
inputs: sources[]
input · retrieve · llm · transform · output
tone-rewrite
The same content, rewritten to a target register — terse, formal, plain-language.
inputs: text, tone
input · llm · output
http-enrich
Look up a record in your own API mid-flow, then reason with it. The SSRF boundary applies.
inputs: record_id
input · http · llm · branch · output
Clone any of these from the app — start free and they are in your account.