Skip to main content
jo runs entirely on your Mac using local AI models. But when you need more powerful reasoning, creative help, or complex analysis, you can optionally call in cloud AI models.

​
How cloud integration works

Cloud integration is opt-in and query-specific. Here’s how it works:
  1. Default: All queries run locally on your Mac
  2. When you need more: Explicitly request a cloud model for that specific conversation
  3. What gets sent: Only the current conversation and relevant context—never your entire indexed data
  4. Pricing: Pay per use, typically $0.01-0.05 per conversation
Privacy promise: Your indexed data (emails, notes, files) stays on your Mac. When you use a cloud model, jo only sends the specific conversation and relevant snippets you’re discussing.

​
Supported cloud models

jo integrates with four major AI providers:

OpenAI GPT-5.2

Best for: General reasoning, coding, creative writingCost: ~$0.02-0.05 per conversationSpeed: Fast (2-5 seconds)

Anthropic Claude

Best for: Long-form analysis, nuanced reasoning, detailed explanationsCost: ~$0.03-0.05 per conversationSpeed: Medium (3-7 seconds)

Google Gemini

Best for: Multimodal tasks, image understanding, research synthesisCost: ~$0.01-0.03 per conversationSpeed: Fast (2-4 seconds)

xAI Grok

Best for: Real-time information, current events, unconventional takesCost: ~$0.02-0.04 per conversationSpeed: Fast (2-5 seconds)

​
Connecting cloud models

You can connect cloud models during initial setup or anytime later:
1

Open cloud settings

In jo, click the gear icon (or press Cmd+,) to open Settings, then navigate to the Cloud Models tab.
2

Choose a provider

Click Connect next to the cloud provider you want to add. Each provider has slightly different setup steps.
3

Add your API key

  1. Visit platform.openai.com/api-keys
  2. Click Create new secret key
  3. Name it “jo” and click Create
  4. Copy the key (starts with sk-)
  5. Paste it into jo’s API Key field
  6. Click Verify & Save
OpenAI requires a paid account with credits. Add at least $10 to start.
4

Test the connection

jo automatically tests the API key when you save it. If successful, you’ll see a green checkmark and your account balance (if available).
Some providers may take a few minutes to activate new API keys.

​
When to use cloud models

Use local AI for most queries. Call in cloud models when you need:
Use case: Multi-step problem solving, strategy, planningExample queries:
  • “Help me prep for this salary negotiation” (after sharing context)
  • “Analyze the pros and cons of these three contractor quotes”
  • “Create a project plan based on my meeting notes”
Why cloud: Local models excel at retrieval and simple queries. Cloud models handle complex synthesis and reasoning.
Use case: Writing, brainstorming, content creationExample queries:
  • “Draft a professional response to this client email”
  • “Help me brainstorm names for this project”
  • “Write a summary of my conference notes”
Why cloud: Cloud models are trained on broader datasets and produce higher-quality creative output.
Use case: Coding, technical explanations, domain expertiseExample queries:
  • “Explain this error message in my code file”
  • “What legal considerations should I know for this contract?”
  • “Help me understand this medical report”
Why cloud: Cloud models have more extensive training in specialized domains.
Use case: Summarizing large amounts of informationExample queries:
  • “Summarize everything I learned at that conference last year”
  • “Create a comprehensive client profile from all my notes and emails with them”
Why cloud: Cloud models can handle longer contexts and produce more coherent long-form content.

​
Using cloud models in queries

There are three ways to invoke a cloud model: Simply mention the model in your query: 
"Call in GPT: help me draft a response to this client email"

​
Method 2: Quick action button

When viewing a local response, click the Enhance with Cloud button that appears above the answer. Select your preferred model from the dropdown.

​
Method 3: Auto-detection (optional)

In Settings > Cloud Models, enable Auto-detect complex queries. jo will suggest using a cloud model when it detects:
  • Requests for creative writing
  • Multi-step reasoning
  • Technical explanations beyond local capability
Auto-detection is disabled by default to prevent unexpected charges. Enable it only if you’re comfortable with automatic cloud usage.

​
Cost transparency

jo shows costs before and after cloud queries:

​
Before sending

When you request a cloud model, jo displays an estimated cost:
“This query will use GPT-5.2. Estimated cost: $0.03” [Confirm] [Cancel]
Click Confirm to proceed or Cancel to stick with local processing.

​
After the response

The actual cost appears at the bottom of each cloud response:
✓ Response generated by GPT-5.2 in 3.2s · Cost: $0.027

​
Monthly tracking

View your cloud usage in Settings > Cloud Models > Usage: Monthly cloud usage breakdown
  • Current month spending
  • Cost per provider
  • Query count and average cost
  • Spending limit warnings

​
Setting spending limits

Prevent surprise charges by setting monthly limits:
1

Open spending settings

Go to Settings > Cloud Models > Spending Limits
2

Set monthly budget

Enter your desired monthly limit (e.g., $10). jo will warn you at 80% and stop using cloud models at 100%.
3

Configure alerts

Choose how to be notified:
  • Banner notification (default)
  • Email alert (enter your email)
  • Block automatically (prevent overspending)
Recommended limits: Most users spend 3−8/month.Startwitha3-8/month. Start with a 10 limit and adjust based on your usage patterns.

​
What gets sent to the cloud

When you use a cloud model, jo sends: ✅ What is sent:
  • Your current query
  • The conversation history (within this session)
  • Relevant snippets from your data (e.g., the specific email you’re asking about)
❌ What is NOT sent:
  • Your complete email archive
  • All your notes or files
  • Your search index
  • Metadata about your data sources
jo sends the minimum context needed to answer your question. For example, if you ask about an email, jo sends that email—not your entire inbox.

​
Privacy and security

​
Provider data policies

Each provider has different data retention policies:
ProviderData RetentionTraining on Your Data
OpenAI GPT30 daysNo (API data not used for training)
Anthropic ClaudeNot retainedNo (API data not used for training)
Google Gemini18 monthsNo (if using Gemini API, not AI Studio)
xAI Grok30 daysNo (API data not used for training)
Important: jo uses the API versions of these models, which have stricter privacy policies than consumer versions (like ChatGPT web interface).

​
Disabling cloud integration

You can disable cloud models entirely:
  1. Go to Settings > Cloud Models
  2. Click Disconnect All
  3. Confirm the action
jo will return to 100% local operation. You can reconnect anytime.

​
Troubleshooting

Solutions:
  • Verify you copied the entire key (they’re long!)
  • Check that the API key is active in the provider’s console
  • Ensure your provider account has credits or an active payment method
  • Try generating a new API key
Possible causes:
  • Network latency (cloud queries require internet)
  • Provider API slowdowns (check status.openai.com, etc.)
  • Large context being sent (reduce conversation length)
Solution: Local queries are always faster. Use cloud only when needed.
Review:
  • Check Settings > Cloud Models > Usage for a breakdown
  • Look for auto-detection settings (may be enabled)
  • Review recent queries in the history panel
Prevent: Set spending limits in Settings > Cloud Models > Spending Limits
Cause: You’ve hit the provider’s usage limits (usually temporary)Solution:
  • Wait a few minutes and try again
  • Check your API tier with the provider (some have stricter limits)
  • Upgrade your provider account for higher limits

​
Next steps

Your first queries

Learn how to ask jo questions effectively

Keyboard shortcuts

Master jo’s keyboard shortcuts for efficiency

Build docs developers (and LLMs) love

Get started for freeTalk to us