Privacy policy

Last updated: April 24, 2026

This page describes what data pudding ("the service", "we", "us") collects, how we use it, and how you can remove it. Plain English; we'll keep the legalese to a minimum.

What we collect

• Account info. Email, name, profile picture — all provided by Clerk when you sign up. We don't see your password.
• Workspace + organization data. The name and members of your workspace, also via Clerk.
• Conversation transcripts. Anything you paste, upload, or sync from Granola/Gong/Zoom. We store the raw text and segments, plus the verbatim quotes Claude extracts.
• Quote tags + embeddings. The classifications we attach to your quotes (pain, objection, feature, competitor) and a 1024-dimensional vector representation we use for semantic search.
• Operational logs. Standard request logs (route, IP, status code) for debugging and rate limiting. Kept ~30 days.

What we don't do

• We don't sell your data.
• We don't train any models on your transcripts. The Anthropic API is configured with log_user_id off and we don't opt into training data sharing.
• We don't share your transcripts across workspaces. Multi-tenancy is enforced at the query layer — every read scopes to your workspace ID.

Subprocessors

• Clerk — auth, sessions, user profiles.
• Neon — the Postgres database where transcripts and quotes live. Encrypted at rest.
• Vercel — hosting and request routing.
• Anthropic — Claude calls for quote extraction and generative drafts.
• Voyage AI — embeddings for semantic search.
• Inngest — async job orchestration for the extraction pipeline.

Data retention + deletion

You can delete every quote, call, and tag in your workspace from Settings → Danger zone → Delete workspace data. The action cascades through every table; nothing remains beyond a brief backup window (~30 days) at our database provider.

Want to fully delete your account? Email zack@trayo.ai and we'll handle it within 7 days.

Contact

Questions or concerns: zack@trayo.ai.