How DraftAria handles your documents.

Aria's features run on Anthropic's Claude models through Anthropic's API. When you use an Aria feature — generating a template, analyzing a document, or chatting — the text you submit (and, for the AI workspace, files you upload) is sent to the model to produce the answer.

Requests to the model are made only from DraftAria's servers. Your browser never holds an AI key, and AI features require you to be signed in and a member of a firm.

Filing checks against court rules ("Ready to File") are rule-based and do not send your document to an AI model.

Documents you create, their version history, analyses, AI workspace messages and uploaded files are stored in your firm's workspace so you can return to them. They are kept until they are deleted.

In transit: the site and its API are served over HTTPS (TLS).

The website is hosted on Vercel. Application data, sign-in and file storage use Supabase.

Firm isolation. Each firm's data is separated in the database using row-level security, so members of one firm cannot read another firm's documents.

Roles. Users have a role within their firm — attorney, paralegal or firm admin. Firm admins manage invitations and members. The shared court catalog can only be changed by DraftAria platform administrators.

Rate limits. AI features are rate-limited per user to prevent abuse.

Audit log. Firm admins can view their firm's audit log. Today it records team invitations; broader coverage (sign-ins, document changes, exports) is on the roadmap.

Not yet available: single sign-on (SAML / Azure AD / Okta) and multi-factor authentication.