My Calendar, Inbox and To-Do List Now Run Themselves (I Handed Them to Claude)

Every Monday I used to spend the first hour of my workweek doing the same three chores by hand. I triaged a Gmail inbox, dragged calendar blocks around, and rebuilt a to-do list in a Google Sheet.

I am an operations consultant in Denver, and those chores were eating close to 100 minutes a week. So for five weeks I handed all three to Claude, connected to the Google account where my work already lives.

By the end, the stack cost me $20 a month and most of that work moved on its own. Not all of it.

I still clicked approve on every email that left my inbox. Here is exactly which parts ran alone, which waited for me, and where the thing broke.

Claude Pro plus Reclaim Lite costs $20 a month

The setup costs $20 a month, and that is the only payment I made.

Claude Pro runs $20 a month on monthly billing, and Reclaim.ai's free Lite plan handled every scheduling job I gave it during the five weeks, so nothing else was charged at setup.

I connected one Google account, which covered Gmail, Calendar and Drive in a single authorization. I added no separate inbox app and no second mailbox. That one account was the whole scope of the test, and it stayed that way.

The catch: the $20 recurs every month; it is not a one-time setup fee. Nothing in my five weeks was billed per email, per task or per calendar move, so the monthly line did not grow as I used it more.

My bill: Claude Pro $20 a month, Reclaim Lite $0, total $20 a month. That gave me one predictable line to check, which is the reason I could run the test for five weeks without opening a spreadsheet to track costs.

Task	Best tool	By hand (minutes)	Claude's own time	Review/edit (minutes)
Inbox triage	Claude with Gmail connector	18	45 sec	4
Calendar replan	Reclaim.ai via Claude	20	1 min	2
Daily to-do list	Claude reading a Google Sheet	25	30 sec	5
Friday review	Claude with Gmail and Calendar	35	2 min	12

The connectors see only the Google account I picked

The connectors reach one Google account: the one I authorized on setup day. My second account and every message in it stayed invisible, including two client threads I had running there.

Anything not shared with the connected account is simply not read, which is the first limit to plan around.

Claude does not get new access. It mirrors the Workspace permissions the connected account already has, so it sees what I see.

A file owned by a colleague and never shared with me stays out of reach, and Claude will not quietly find a copy somewhere else.

Answers come back with citations: the emails, calendar events and documents it used, with links to the originals where the tool could produce one.

On each summary I open one cited email to confirm the source. That single check costs me a few seconds and replaces re-reading the thread.

The caveat: a citation proves where a line came from, not that the line is right.

If the relevant message sits in the account I never connected, the summary reads fine and is incomplete, which is why I compare the cited list against the threads I know matter.

The human click stays at that source check. I act on a summary after one link, not after reading six messages.

Gmail sees attachment names, not contents

Even with the right account connected, Claude's Gmail link hands over the envelope and not the paper inside. Attachment contents do not come through the connector at all.

What it returns is metadata, so I see that a file called "MSA_v3_final.pdf" arrived and who sent it, and nothing about the terms or the total.

Filters are the second blind spot. Some advanced Gmail filters are not supported through the connector, so a rule I set years ago can quietly keep a message out of the summary.

I now check that any client domain has no filter hiding it before I trust a quiet inbox.

A contract landed on a Tuesday with a renewal clause inside. I asked Claude to pull the clause, and the reply gave me the filename and sender. No number, no term.

I saved the PDF to Google Drive and asked again there, and it read the page and flagged a renewal price that had gone up.

That detour is what caught the bill, and it is why invoice and contract PDFs now go to Drive first. The click that stays mine is confirming the Drive copy is current, not scrolling a thread.

Inbox triage runs itself; sending waits for me

Inbox triage is the half of email that runs on its own.

Every weekday at 7am I ask Claude to go through unread threads in the connected Gmail account, and it comes back with a summary that cites each email it used, with a link to the original.

It also drafts the replies. Those drafts land unsent. On my Pro setup, send, reply and forward all require my explicit approval. Team and Enterprise owners can let some actions run without asking each time; I left that alone.

That rule is the whole safety margin. A wrong draft costs me a correction before it goes out, not an apology to a client.

The catch is that triage is only as good as the citations, so I open one cited thread to check the summary before I touch a draft.

By hand, the 7am pass took me 18 minutes. Claude's summary takes about 45 seconds to appear, and my review and edits take 4 minutes. The rest of the time comes back to me.

Calendar re-planning is the part that truly runs

Calendar blocks are where work moved without my hand on it. I connected Reclaim.ai, a scheduling app, to Claude through Reclaim's hosted MCP server. That's the connector piece that lets Claude hand scheduling requests to Reclaim instead of retyping them.

Tuesday at 9:40 a.m. a client moved a call into a block I had set aside for proposal writing. Reclaim found the collision, shifted the writing block to Thursday and pushed the change to my Google Calendar before I opened a calendar tab.

The free Lite plan caps scheduling at a one-week range, with 1 Calendar Sync and 1 Scheduling Link. Starter is $10 per seat per month billed annually. Book a task six days out and it schedules; eight days out, the free horizon runs out.

By hand, replanning that collision costs me about 20 minutes of dragging blocks and checking conflicts. Reclaim returned a plan in about 1 minute. My check took 2 minutes, reading the new blocks against two meetings it could not move.

Time check: 20 minutes by hand, 1 minute from Reclaim, 2 minutes of my review. That 2-minute review is the only click left in the loop.

To-do list: a daily brief, not a robot

The task list is the opposite case. Nothing writes back to my Google Sheet, so the output is a brief I retype. Each morning at 7:15 I open a chat and ask Claude to read that sheet through the Drive connector.

It returns three priorities with time estimates in about 30 seconds, ranked by deadline and time. Last Tuesday it put the overdue invoice first and pushed a low-stakes call to the afternoon.

Then the catch: Claude reads the sheet and never reorders it. I paste the three lines into the sheet myself, and that paste is part of the review, not a formality. Twice I caught an estimate that ignored a meeting block.

Planning by hand used to cost 25 minutes at my desk with coffee going cold. Now it is 30 seconds from Claude and 5 minutes of my correction pass. The list is ready before my first meeting, sorted the way I actually want it.

Friday review catches what the daily loop misses

Loose ends pile up through the week, and Friday's pass is where I settle them before Monday. I ask Claude to list what is still unresolved: email threads with no reply from me, calendar blocks that moved, and tasks that slipped on the sheet.

By hand that sweep takes me 35 minutes of scrolling three places. Claude's first pass takes about 2 minutes and comes back with the three lists side by side.

That first pass is a draft, not the decision. My correction and decision pass takes 12 minutes, and it has caught at least one thread I forgot every week so far.

One client thread surfaced only because it sat in unresolved email. The daily loop had already moved past that address days earlier.

The 12 minutes is what makes Monday's list trustworthy. The click stays on the one call Claude cannot make: what to carry forward and what to drop.

Heavy days hit the five-hour window

When Claude stopped answering mid-afternoon, the clock was the reason, not a bug. There is no fixed message count. Usage depends on how long the conversations are, their complexity, and the features you switch on.

Limits reset on a rolling five-hour session, and paid plans add weekly caps on top. On Pro I get at least 5x the Free usage in each session. A long agent run still spends from that same window.

I learned it the hard way. I let a big research ask eat an afternoon before I had touched triage, and the cheap work then had to wait.

Now triage goes first. The 45-second inbox pass gets done while the window is fresh, and the expensive ask gets whatever is left.

The order is the whole fix. Put the short, certain work ahead of the long run and a heavy day no longer costs you the morning's easy wins.

The privacy line is no secrets, even with connectors

Passwords, client files, and medical or financial records never go into the chat, whether or not a connector is attached. A connector widens what Claude can read, and it also widens what you might be tempted to hand it.

The training lines are not the same. Anthropic says it does not train on Gmail, Drive or Calendar connector data. On free and personal plans, ordinary chats may still be used to improve models unless you switch that setting off.

So I work on a redacted copy. The secret stays in the source app, and Claude reads the connector version of the source rather than a pasted original.

I keep a scratch document with account numbers and names blacked out. Claude summarizes that copy. The summary is still useful, and nothing sensitive ever leaves the app it lives in.

That is the whole rule. Useful summaries survive redaction, and the human click stays on what leaves the source app. On a personal plan, check that setting before you trust the chat.

Frequently Asked Questions

Do I need to know how to code to set this up?

No. The setup is connecting a Google account inside Claude's settings and, if you want the calendar re-planning, linking Reclaim through its hosted server. You never touch a script or a terminal.

Will Claude send emails on its own?

On a personal Pro plan, no. Sending, replying and forwarding require explicit approval by default. Anthropic's documentation says Team and Enterprise owners can allow some actions to run without asking each time, so check that setting if you are on a work plan.

What does the whole stack actually cost?

In my test, $20 a month for Claude Pro was the only charge, with Reclaim's free Lite plan covering scheduling. Reclaim Starter at $10 per seat per month billed annually is the upgrade if you need a scheduling range longer than one week. Check both pricing pages before you commit.

Why could it not read my email attachments?

Claude's Gmail connector works from metadata, so it sees a filename but not the page inside it. Save the PDF to Google Drive and ask Claude to read the Drive copy instead, then confirm the figures yourself.

Is my email used to train the model?

Anthropic says it does not train on Gmail, Drive or Calendar data coming through the connector. Chats on free and personal plans may still be used to improve models unless you switch that off in settings.

What happens when I hit the usage limit?

Claude stops answering until the rolling five-hour session resets, and paid plans also carry weekly caps. Do the short, cheap tasks first so the useful work is done before a long research request spends the window.

Worth $20 if you will click approve

Set it up if your calendar and inbox already live in Google and you will read each draft before it goes. I would run the same $20 stack again after five weeks, for triage and replanning, with the send button still mine.

Skip it if mail has to leave without your approval, or if your workload arrives mostly as Gmail attachments. The connector sees the filename only, so a contract summary stops at the attachment name until you save a copy to Drive.

If a week of schedule is not enough, budget for the Reclaim upgrade; Starter is $10 per seat per month billed annually. Lite's one-week horizon and single sync are the free ceiling, not a starting point you can stretch.

Start with one week: Claude Pro plus Reclaim's free Lite plan, one Google account, four tasks. Inbox triage, a calendar replan, the daily to-do brief and the Friday review are the test. The approval click is the verdict.

The approval click is the product, not the compromise

If you live in Google Workspace and want your calendar and inbox triaged while you keep the send button, the $20 stack of Claude Pro and Reclaim's free Lite plan is worth one week of your time.

Skip it if you need unattended sending or your work arrives mainly as Gmail attachments, because those are the two places this setup fails first. Spend one week on the stack before adding Reclaim Starter, and count the clicks you keep making.

The chores moved. The judgment did not, and that is precisely why I would set it up again.