🦞 4minAI.com
1 / 12
Day 8 of 28 Β· OpenClaw Challenge

Connecting to Email

Welcome to Week 2. Last week you learned the fundamentals and got OpenClaw running. This week, you connect it to your real world.

We're starting with the tool that will save you the most time: email.

Why email first?

The average professional spends 28% of their workday on email. That's 11 hours a week reading, sorting, drafting, and sending.

What if your agent handled 80% of that? Not by ignoring your email β€” by actually reading it, sorting it, drafting replies, and letting you review with a quick "looks good, send it."

Knowledge Check
What percentage of the average workday is spent on email?
A
About 10%
B
About 28% β€” roughly 11 hours a week
C
About 50%
D
Less than 5%
Studies consistently show email eats about 28% of the average knowledge worker's day. That's over 11 hours a week β€” time your agent can give back to you.

Connecting Gmail or Outlook

OpenClaw connects to your email through a secure API. You authorize it once β€” just like when you connect a new app to your Google account β€” and the agent can read and send email on your behalf.

Important: your emails stay on your machine. OpenClaw pulls them locally and processes them locally. Nothing is stored on any third-party server.

🦞
OpenClaw 🦞
online
I just connected my Gmail. What can you do with it?
Now that I have email access, I can: πŸ“₯ **Read** β€” scan your inbox, search for specific emails, find attachments πŸ“‚ **Sort** β€” categorize by importance, sender, or topic ✏️ **Draft** β€” write replies based on context and your tone πŸ“€ **Send** β€” send emails on your behalf (always with your approval first) πŸ—‘οΈ **Clean** β€” archive old threads, unsubscribe from newsletters Want me to start with an inbox scan?
↻ Replay conversation
Knowledge Check
When OpenClaw reads your email, where does the processing happen?
A
In your web browser
B
On OpenClaw's cloud servers
C
On Google's AI servers
D
Locally on your own machine
OpenClaw processes everything locally. Your emails are pulled to your machine and analyzed there. They're never sent to OpenClaw's servers β€” because OpenClaw doesn't have servers. It's all yours.

Your first inbox triage

Let's see what a morning inbox triage looks like with OpenClaw:

🦞
OpenClaw 🦞
online
Good morning. What's in my inbox?
Scanning your inbox...
Morning! You have 34 new emails since last night. Here's the breakdown: ⭐ **Need your reply (4)**: 1. Sarah Chen β€” "Q1 budget approval needed" (urgent) 2. Client (Acme Corp) β€” "Proposal feedback" 3. Your manager β€” "Friday demo prep" 4. HR β€” "Benefits enrollment deadline tomorrow" πŸ“‚ **FYI only (8)**: Team updates, meeting notes, status reports β€” archived to 'Read Later' πŸ—‘οΈ **Noise (22)**: Newsletters, promotions, automated notifications β€” moved to trash I've drafted replies for all 4 urgent emails. Want to review them?
Show me the draft for Sarah's email.
Here's my draft reply to Sarah: --- Hi Sarah, Approved. The Q1 numbers look solid β€” nice work on the 15% reduction in cloud costs. One note: can we revisit the contractor budget in our Wednesday sync? I think there's room to reallocate. Thanks, [Your name] --- I based this on the actual budget spreadsheet she attached and your previous emails about cost reduction. Edit or say 'send' when ready.
↻ Replay conversation
Knowledge Check
How did the agent know to mention "15% reduction in cloud costs" in the draft?
A
It made it up to sound impressive
B
It searched Google for average cost reductions
C
It copied text from a previous email
D
It read the budget spreadsheet Sarah attached and extracted the key data
The agent read Sarah's attachment, analyzed the budget data, and pulled out the most relevant number. It didn't just draft a generic reply β€” it drafted an informed reply based on the actual content.

The trust model

You might be nervous about an AI sending emails on your behalf. That's smart. OpenClaw is designed with a review-first approach:

The agent drafts emails but doesn't send them without your say-so. You review, edit if needed, and then approve. Over time, as you build trust, you can adjust the permissions β€” letting the agent auto-send certain types of emails (like meeting confirmations) while still requiring approval for others (like client communications).

You're always in control.

Final Check
What's the default behavior when OpenClaw drafts an email?
A
It can only draft β€” it can never send
B
It sends immediately without asking
C
It sends but lets you undo within 30 seconds
D
It saves the draft and waits for your review and approval before sending
Review-first is the default. The agent drafts, you approve. As trust builds, you can gradually give it more autonomy for low-risk emails. But you're always in control of that dial.
πŸ“§
Day 8 Complete
"Your agent just took over your inbox. 28% of your workday β€” reclaimed."
Tomorrow β€” Day 9
Connecting to Your Calendar
Email is handled β€” now let's give your agent eyes on your schedule.
πŸ”₯1
1 day streak!