Hello, world. This is agentl. Describe a data job in one YAML file. agentl reads your files and APIs, runs LLM steps that return typed fields, checks the data and writes it where you need it.

Built so LLM pipelines don't burn money: • per-record cache: run it again, you don't pay twice • budget caps stop a run before it overspends • cheap model first, strong model only when needed • resumes where a failed run stopped