Frequently Asked Questions

Common questions about InboxZebra’s privacy, functionality, and features.

Privacy & Data Security

Do you train on my email data?

Absolutely not. Your email data is never used to train our models or improve InboxZebra for other users.

Here’s exactly what happens with your data:

Local Learning System (Improves YOUR Results Only):

• When you manually correct a categorization, InboxZebra stores a sanitized example (subject line, sender, 100-character snippet) locally on your Mac
• These examples are used ONLY to improve categorization accuracy for YOUR account
• Stored in ~/Library/Preferences/com.inboxzebra.app.plist on your Mac
• Never transmitted to our servers or anyone else
• You can export/delete this data anytime in Settings → Data

What We Do NOT Do:

• ❌ Collect your email content
• ❌ Send your data to central servers
• ❌ Aggregate data across users
• ❌ Train models using your emails
• ❌ Share your data with third parties
• ❌ Use your data for any purpose other than providing the service to YOU

How Local AI Works:

• Default processing uses Ollama running locally on your Mac
• The AI model (e.g., qwen2.5:7b) runs entirely on your device
• Your emails never leave your computer for categorization
• No internet connection required for AI processing

Optional Cloud AI (Business Tier):

• You can optionally use Google Gemini with your own API key
• Data goes directly from your Mac to your Gemini account (not through us)
• You control the API key and can revoke access anytime
• We never see or store what’s sent to Gemini
• Completely optional - local processing is the default

Bottom Line: Your data belongs to you. We never see it, store it centrally, or use it to train anything. InboxZebra is designed for privacy from the ground up.

What data does InboxZebra collect?

On Your Mac (Local Storage):

• UserDefaults: Categories, settings, account metadata (email addresses, display names), learning examples (sanitized), custom instructions
• Keychain: OAuth access and refresh tokens (encrypted by macOS)
• Core Data: Email metadata (subject, sender, dates) and encrypted email bodies (AES-256-GCM)

On Our Servers:

• Nothing. We don’t have servers that store your email data.
• License validation happens via Paddle (our payment processor) but doesn’t involve email content.

Third-Party Services:

• Gmail API: For fetching/updating Gmail emails (direct connection from your Mac to Google)
• Microsoft Graph API: For fetching/updating Office 365 emails (direct connection from your Mac to Microsoft)
• Ollama: Runs locally on your Mac (no external connection)
• Gemini API (optional): If you enable cloud AI with your own API key, requests go directly to Google

Can you see my emails?

No. We have no access to your emails. Here’s why:

1. OAuth Authentication: You authenticate directly with Gmail/Microsoft. We never see your email password.
2. Direct API Calls: Your Mac communicates directly with Gmail/Microsoft servers. No intermediary.
3. Local Processing: All AI categorization happens on your Mac via Ollama (or your own Gemini API key if you enable cloud AI).
4. No Central Servers: We don’t operate servers that handle your email data.

The only way we could see your emails is if you explicitly sent them to us (e.g., for support troubleshooting), which we would never ask you to do.

Do you use analytics or telemetry?

No. InboxZebra has zero analytics or telemetry.

• No tracking pixels
• No usage analytics
• No crash reporting to external services
• No “phone home” functionality

The only network requests InboxZebra makes are:

1. OAuth authentication with Gmail/Microsoft
2. Fetching/updating emails via Gmail/Microsoft APIs
3. Optional: Gemini API calls (if you enable cloud AI with your own API key)
4. License validation (via Paddle, our payment processor)

How is my data encrypted?

At Rest (On Your Mac):

• Email bodies: AES-256-GCM encryption in Core Data
• OAuth tokens: Stored in macOS Keychain (encrypted by macOS with hardware-backed keys)
• Settings/categories: Stored in UserDefaults (standard macOS file permissions)

In Transit:

• All API calls use HTTPS (TLS 1.3)
• OAuth tokens transmitted only over secure connections
• Email content never transmitted except to Gmail/Microsoft servers (over HTTPS)

Access Control:

• Touch ID/password required for viewing email bodies
• Keychain access restricted to InboxZebra app only
• macOS sandboxing limits file system access

Features & Functionality

What’s the difference between categories and labels?

Categories:

• Primary classification (e.g., Work, Personal, Newsletters)
• Each email belongs to exactly ONE category
• Synced to Gmail/Office 365 as folders/labels
• Used by AI for automatic categorization
• Can have descriptions to guide the AI

Secondary Labels:

• Additional metadata (e.g., Finance, Travel, Urgent)
• Each email can have MULTIPLE labels
• Local only (not synced to email providers)
• Used for filtering and organization
• Support custom colors and icons (Pro tier)

Example: An email might be categorized as “Work” with secondary labels “Finance” and “Urgent”.

How does the learning system work?

InboxZebra learns from your corrections to improve accuracy over time:

1. Initial Categorization: AI analyzes email content and assigns a category
2. Manual Correction: You right-click and recategorize if it’s wrong
3. Learning: InboxZebra stores a sanitized example (subject, sender, snippet)
4. Improved Prompts: Future categorization uses your examples for better accuracy
5. Local Only: Learning data stored only on your Mac, never shared

Priority:

• Manual corrections (marked with ✓) are highest priority
• Recent automatic categorizations fill remaining slots
• Example limits match your plan:
  • Free tier: limited to 3 saved examples per category
  • Pro/Business tiers: full access with 10+ saved examples per category

Privacy:

• Only 100 characters of email body stored
• Sanitized snippets stored in UserDefaults (not encrypted; full email bodies are stored encrypted in Core Data)
• Can be exported/deleted in Settings → Data

Can I use InboxZebra offline?

Partially.

Offline Capabilities:

• ✅ View previously fetched emails
• ✅ Categorize emails (if Ollama is running locally)
• ✅ Manage categories and labels
• ✅ Search cached emails

Requires Internet:

• ❌ Fetching new emails (requires Gmail/Microsoft API access)
• ❌ Applying categories/labels to email provider (requires API access)
• ❌ OAuth authentication (requires internet)
• ❌ License validation (requires internet for initial activation)

What’s the difference between Pro and Business tiers?

14-day free trial includes all Pro + Business features.

Why would I use cloud AI (Gemini) instead of local AI (Ollama)?

Local AI (Ollama) - Default:

• Pros: Completely private, no API costs, works offline, no limits
• Cons: Requires 16GB+ RAM, slower (2-5 seconds per email), uses CPU/GPU

Cloud AI (Gemini) - Optional (Business tier):

• Pros: Faster (0.5-1 second per email), no local hardware requirements, parallel processing
• Cons: Requires your own API key, costs ~$1/month for 10k emails, internet required, data processed via Google

When to use Gemini:

• Processing thousands of emails regularly
• Limited local hardware (8GB RAM Macs)
• Want fastest possible categorization
• Comfortable with Google seeing email subjects/snippets

You provide your own API key, so costs and data go to your Google account, not through us.

Technical Questions

What AI models are supported?

Local AI (Ollama):

• Recommended: qwen2.5:7b (best balance of speed/accuracy)
• Fastest: phi3:mini (good for 16GB RAM)
• Highest Quality: gemma2:9b (requires 32GB+ RAM)
• Any Ollama model can be configured in Settings

Cloud AI (Gemini - Business tier):

• gemini-2.5-flash (fast, cost-effective)
• Automatic model selection by Google

What email providers are supported?

Currently Supported:

• ✅ Office 365 (Outlook.com, Microsoft 365, Exchange Online)
• ✅ Gmail (personal and Google Workspace accounts)

Shared Mailboxes:

• ✅ Office 365 shared mailboxes (Business tier)
• ❌ Gmail shared/delegated mailboxes (not yet supported)

Not Supported:

• ❌ IMAP/POP3 accounts
• ❌ Yahoo Mail
• ❌ ProtonMail
• ❌ Other providers

We focus on Office 365 and Gmail because they represent 80%+ of business/personal email and have robust OAuth2 + modern APIs.

What are the system requirements?

Minimum:

• macOS 13.0 or later
• 16GB RAM (for 7B-8B models)
• 10GB free storage (for model + app)
• Internet connection (for fetching emails)
• Ollama installed (for local AI)

Recommended:

• macOS 14.0 or later
• 32GB RAM (for larger/faster models)
• Apple Silicon (M1/M2/M3) for best performance
• 50GB free storage (for multiple models)

Cloud AI (Gemini) Requirements:

• 8GB RAM minimum (no local AI needed)
• Stable internet connection
• Google Gemini API key (free tier available)

How do I backup my data?

Export Settings:

1. InboxZebra → Settings (Cmd+,)
2. Data tab → “Export Data”
3. Save JSON file to safe location

What’s Included:

• ✅ Categories and descriptions
• ✅ Learning examples
• ✅ Account metadata (email addresses, names)
• ✅ Custom labels
• ✅ Preferred senders
• ✅ Custom instructions
• ✅ LLM settings

What’s NOT Included:

• ❌ OAuth tokens (must re-authenticate after import)
• ❌ Actual email content
• ❌ Cached emails

Restore:

1. InboxZebra → Settings (Cmd+,)
2. Data tab → “Import Data”
3. Select exported JSON file
4. Choose “Replace All” or “Merge”
5. Re-authenticate accounts

Subscription & Licensing

How does the free trial work?

14-Day Free Trial:

• Includes ALL Pro + Business features
• No credit card required
• Starts automatically on first launch
• Full access to unlimited accounts, cloud AI, custom labels
• After 14 days: downgrade to Free tier or purchase Pro/Business

Free Tier After Trial:

• 1 email account
• 500 emails/month
• Basic features (categories, learning, newsletter management)
• Can upgrade anytime

How do I activate my license?

After Purchase:

1. You’ll receive a license key via email from Paddle
2. Open InboxZebra → Settings (Cmd+,)
3. Subscription tab
4. Enter your license key and email (exactly as in purchase confirmation)
5. Click “Activate License”
6. Subscription activates immediately

Troubleshooting:

• Ensure key is entered exactly (copy/paste recommended)
• Check email matches purchase email
• Restart InboxZebra after activation
• Contact support if issues persist

Can I use InboxZebra on multiple Macs?

License Terms:

• Pro: 1 license = 1 active installation
• Business: 1 license = 1 active installation

Multiple Macs:

• You can install on multiple Macs but only one can be active with the license key
• Deactivate on one Mac before activating on another
• Or purchase additional licenses for each Mac

Family/Team:

• Each user needs their own license
• Business tier supports team discounts (contact sales)

What happens if my subscription expires?

After Expiration:

• App continues working
• Features revert to Free tier:
  • 1 account only (others disabled)
  • 500 emails/month limit
  • Custom labels disabled (but preserved)
  • Cloud AI disabled
  • Shared mailboxes disabled

Your Data:

• All data preserved (categories, learning, emails)
• Can still view previously categorized emails
• Can export data anytime

To Restore:

• Renew subscription
• All Pro/Business features re-enabled immediately
• Data/settings restored automatically

Troubleshooting

Why is categorization slow?

Local AI (Ollama):

• Sequential processing: 2-5 seconds per email
• Large models (70B+) can be 10-30 seconds per email
• Solution: Use faster model (qwen2.5:7b or phi3:mini)
• Or: Upgrade to Business tier and use cloud AI (Gemini)

Cloud AI (Gemini):

• Parallel processing: 10-30 emails at once
• 0.5-1 second per email
• Rate limited on Gemini’s free tier
• Solution: Increase monthly quota in Settings

General:

• Close other memory-intensive apps
• Reduce “Max Emails to Fetch” in Settings
• Process in smaller date ranges

Why are some emails miscategorized?

Common Causes:

1. Vague category descriptions: Add detailed descriptions in Settings → Categories
2. Insufficient examples: Manually correct 5-10 emails per category to train the system
3. Conflicting custom rules: Check Settings → LLM → Global Custom Rules
4. Wrong model: Some models are better at instruction-following (qwen2.5:7b recommended)

Solutions:

• Right-click miscategorized emails → Recategorize (InboxZebra learns from this)
• Add per-account context in Settings → Accounts → (account) → Custom Instructions
• Use more specific category names (e.g., “Client Projects” instead of “Work”)
• Try a different model in Settings → LLM

How do I report a bug?

1. Check logs: Console.app → search subsystem:com.inboxzebra.app
2. Export logs:

   log show --predicate 'subsystem == "com.inboxzebra.app"' --last 1h > logs.txt
   

3. Collect info:
   • macOS version
   • InboxZebra version (InboxZebra → About)
   • Steps to reproduce
   • Expected vs actual behavior
   • Relevant log excerpts (redact sensitive info)
4. Contact support: https://inboxzebra.com/support

Privacy Note: When sharing logs, redact email addresses, tokens, and personal content.

Getting Help

Documentation:

• Getting Started
• Troubleshooting Guide
• Full Documentation

Support:

• Free tier: Community support via documentation
• Pro tier: Email support (response within 48 hours)
• Business tier: Priority email support (response within 24 hours)

Contact:

• Website: https://inboxzebra.com/support
• Check for updates: https://inboxzebra.com/download

Still have questions? Visit https://inboxzebra.com/support or check our full documentation at https://inboxzebra.com/docs