Choose a drafting mode, then draft step by step

Our production product runs live with paying advocates (currently on AWS). For this immersion we deployed a dedicated, open-access instance on GCP so you can run the full flow without an account. It's the real pipeline, not a mock. Built on Google's ADK — a sequential pipeline orchestrating the drafting stages, parallel agents running each petition track simultaneously, and shared session state keeping names, dates and statutory sections consistent across every document. Stateful where the flow needs memory across steps, stateless where it doesn't. Deployed through Antigravity onto Cloud Run. Pick a sample case and hit Generate to watch the agents classify the matter, draft it, and validate each section against the statute, then produce a court-ready draft. This GCP build mirrors our production architecture, and the migration path off AWS is straightforward — the agent layer is model- and infra-agnostic by design.