Your data, explained.

Everything Draftroom accesses, stores, and how to delete it. No legal jargon.

What we access vs. what we store

Feature

What we access

What we store

What we discard

Sent Emails

Full email body (up to 1,000 emails, last 12 months)

Voice patterns: tone, vocab, style per contact

Email text after analysis

Inbox

Sender, subject, preview (~1,000 chars)

Classification: needs reply? + confidence

Email bodies, attachments

Original Emails

Full body fetched live from Gmail when you review a draft

Nothing — fetched on demand, never stored

Entire email after you close the view

Reply History

Reply status per email (12 months, up to 1,000)

Reply rate and patterns per contact

Email content

Contact Profiles

Derived from sent email analysis

Tone, opening/closing style per person

Source email text

Calendar

Last 90 days + 30 days ahead, up to 500 events

Title, time, attendees, summary

Raw description after summarization

Drive

Up to 500 documents from the last year

Title, AI summary, semantic embedding

Full document text after summarization

Drafts

We generate these

Full draft, confidence, reasoning

You control: send, edit, dismiss, delete

Learnings

Your edits to drafts

Structured rule: what changed and why

Permanent until you delete

Knowledge Base

Content you paste or upload

Extracted insights + embeddings

Source text after extraction

What we never touch

Email attachments — never downloaded, never processed
Stored inbound email bodies — original emails are fetched live from Gmail when you review a draft, then discarded
Browsing history or device fingerprints — no analytics, no tracking pixels
Payment details — Stripe handles all billing. We never see card numbers.
Contacts outside your email — we only learn about people who email you

Third-party services

Service

What they receive

Retention

Anthropic (Claude)

Email text during analysis

Zero retention

Google APIs

Read-only access via OAuth

Google's policy

Voyage AI

Text for semantic embeddings

API-based, no retention

Supabase

All app data (encrypted at rest)

Standard hosting

Stripe

Subscription metadata only

PCI-DSS Level 1

Your controls

Export everything

Download all your data as a ZIP file with JSON.

Settings → Account → Export

Delete everything

Cascading delete. Permanent within 30 days.

Settings → Account → Delete

Disconnect Gmail

Revokes access. Existing drafts stay until you delete them.

Settings → Gmail → Disconnect

Delete learnings

Reset everything Draftroom has learned from your edits.

Settings → Learnings → Reset

The shape of your data

For the technically curious — here’s what your data actually looks like in our system.

Voice Understanding

{
  "global_patterns": {
    "tone": "friendly",
    "opening_style": "greeting",
    "closing_style": "thanks",
    "avg_sentence_count": 4,
    "key_patterns": ["uses dashes for asides", "starts with 'Hey'"]
  },
  "vocabulary_patterns": {
    "preferred_phrases": ["sounds good", "let me know"],
    "avoided_phrases": ["per my last email"],
    "exclamation_frequency": 0.3,
    "emoji_frequency": 0.1
  }
}

A Learning

{
  "type": "voice",
  "source_action": "edit",
  "description": "When replying to Sarah, use 'Hi Sarah' not 'Hey Sarah'",
  "applies_to_sender": "sarah@company.com",
  "weight": 1.5
}

A Contact Profile

{
  "contact_email": "boss@company.com",
  "tone_override": "professional",
  "reply_rate": 0.95,
  "opening_style": "greeting",
  "emails_analyzed": 47
}