Product
How Supra AI works
Supra AI is organized around the way legal work actually moves: local model control, global questions, matter workspaces, documents, sources, verification, and drafts.
Step 01
Configure local models
Assign separate local MLX models for legal reasoning, high-quality research, drafting, and critique, then load the runtime on device.
Step 02
Start in global chat
Open to a fresh legal prompt set, search prior global chats by title, delete stale chats, or move a chat into a matter when it becomes case-specific.
Step 03
Open a matter
Keep matter chats, research sessions, authorities, documents, structured outputs, and audit history in one workspace.
Step 04
Import documents
Bring in matter files for OCR, extraction, chunking, local embedding search, cited Q&A, and fact chronologies.
Step 05
Review grounded answers
Run CourtListener-backed legal research or document Q&A, then inspect the sources and verification warnings before relying on the work.
Step 06
Draft, critique, or export
Turn reviewed research and document context into attorney-edited drafts, critique passes, chronologies, and exportable work product.