How I Used Claude Code to Unsubscribe from 130 Newsletters in 30 Minutes

I never get around to cleaning out my inbox because there are so many recurring marketing emails I have accumulated over the last 15 years with this account. Smug Inbox Zero proponents would look down their nose at my unread count. The problem had grown so large that it felt unaddressable. Until now.

According to an Exchange message trace, I receive over 90 emails a day, compared to around 20 for our engineers whose inboxes are mostly support tickets and SharePoint notifications.

At GCIT, we use Claude Code every day for development, building our AI coworker, creating MCP tooling, and automating client infrastructure. But last night I had the thought of setting it loose on a different problem: my inbox. The idea was simple. Register a short-lived Azure AD app with read-only access and the ability to create inbox rules, point Claude Code at the Graph API, and let it find every sender with an unsubscribe method so I could triage the deluge of communications I receive each day.

Using Claude Code as my AI assistant, the Microsoft Graph API, and Exchange Online, I triaged my entire mailbox in a single session. The result: 2,000 emails scanned across 90 days, 294 newsletter senders identified, 130 automatically unsubscribed, and 107 more filtered into a dedicated folder with inbox rules. The whole exercise took about 30 minutes, compared to the six to eight hours it would have taken to do manually.

Here is exactly how it worked, what the data looked like, and why this matters for anyone running a business on Microsoft 365.

The Problem Every Business Owner Ignores

The average office worker receives 121 emails per day, according to a 2024 report from Radicati Group. A significant portion of that volume comes from marketing emails, newsletters, and automated notifications, not from colleagues or clients.

In my case, I scanned 2,000 emails from a 90-day window and found that 400 of them (20%) came from newsletter or marketing senders that had unsubscribe mechanisms. That is roughly 4-5 marketing emails per day landing in the inbox alongside legitimate business communication.

The challenge is that unsubscribing one at a time is tedious. Each email requires scrolling to the footer, clicking a tiny link, waiting for a page to load, sometimes confirming on a second page, and occasionally logging into an account. Multiply that by 130 senders and you are looking at hours of manual clicking.

How Claude Code Made This Possible

This is where it gets interesting. I did not write every line of code by hand. I described what I wanted to Claude Code (Anthropic’s AI coding assistant that runs in your terminal), and it built the tools for me in real time.

Here is how the actual conversation started:

Within minutes I had a working triage tool and a ranked report of 117 newsletter senders. From there, the conversation was just me making fast decisions while Claude Code handled the execution:

That was the whole workflow. I described the problem, Claude Code built the tools, and we went through every sender together. My input for most of the session was just “1, 2, or 3” for each sender. Claude Code handled the Azure app registration, the Graph API calls, the MIME header decoding, the one-click unsubscribe requests, the Exchange inbox rules, and even a Playwright browser script for Temu. The whole thing took 30 minutes.

That is the key insight about AI tools like Claude Code for business owners. You do not need to be a developer to get developer-level outcomes. You need to be able to describe the problem clearly, and the AI handles the implementation. The conversation transcripts above are real, and they show how natural the interaction is. No coding syntax, no documentation reading, just plain English instructions and quick decisions.

What We Built: The Technical Approach

The tool connects to a Microsoft 365 mailbox using application permissions via the Graph API. It does three things:

1. Scans the mailbox for emails with unsubscribe methods (List-Unsubscribe headers or body links)
2. Reports on sender frequency, unsubscribe method types, and sample subjects
3. Executes unsubscribes automatically using RFC 8058 one-click POST requests

One technical detail worth noting: many email providers encode their List-Unsubscribe URLs using MIME Q-encoding. Our script needed to decode these before extracting the URLs, a step that is easy to miss and causes silent failures. Claude Code identified this edge case after the first few unsubscribes failed quietly, and added the decoding logic automatically.

The Data: What 90 Days of Inbox Noise Looks Like

The initial 30-day scan found 117 unique newsletter senders across 500 messages. When I expanded to 90 days and 2,000 messages, that number jumped to 294 unique senders. The extended scan caught 177 additional senders that would have been completely missed in a shorter window, seasonal promotions, monthly newsletters, and low-frequency marketing.

The breakdown tells an interesting story about email sender compliance. The vast majority of reputable senders now support RFC 8058 one-click unsubscribe, which means a tool can send a single POST request and you are done. No clicking, no confirmation pages, no logging in. Our script prioritised that method, falling back to HTTP GET, then flagging mailto and body-link senders for manual action.

The Triage: Not Everything Gets Unsubscribed

Rather than blindly unsubscribing from everything, I went through every sender and categorised them into three actions:

Across all 294 senders, the categories broke down roughly as you would expect. Retail and shopping was the biggest chunk (28%), dominated by the usual suspects like David Jones, Amazon, and Starbucks. Tech and software came next (18%), then travel and loyalty programs (14%), news and media (12%), food delivery apps like Temu and DoorDash (10%), cold outreach (8%), and local events and community newsletters (10%).

The Temu Problem: When Senders Make It Hard

Temu was a notable outlier. Despite sending 6 emails in 30 days, it did not include a List-Unsubscribe header. The only unsubscribe option was a small link buried in the email body, wrapped in a Microsoft SafeLinks URL. No API call was going to handle this one.

I suggested Claude Code use Playwright to deal with it. Playwright is an open-source browser automation tool from Microsoft that lets you script a headless browser to navigate pages, click buttons, and take screenshots, the kind of thing QA teams use for testing web apps. Claude Code wrote a Playwright script that launched a headless Chromium browser, navigated to the Temu unsubscribe page, located the “Unsubscribe” button, clicked it, and captured a screenshot of the confirmation. The whole thing took about 30 seconds to write and run.

This highlights a gap in email sender compliance. RFC 8058 has been widely adopted by reputable senders, but some high-volume marketers still rely on body links that are harder for users (and tools) to find and act on. The workaround? AI writes you a browser automation script in 30 seconds.

Creating Inbox Rules at Scale with Exchange Online

For the 107 sender domains I wanted to filter rather than unsubscribe from, I used Exchange Online cmdlets to create inbox rules. First, a “Newsletters” subfolder under Inbox. Then rules using FromAddressContainsWords to match sender domains and automatically move those emails to the folder, marked as read.

One constraint: Exchange limits FromAddressContainsWords to 255 characters. Claude Code calculated the character counts and split the domains across 10 rules to stay within the limit. That kind of tedious detail is exactly where AI assistants shine, handling the fiddly constraints so you can focus on the decisions.

What Does This Mean for Your Business?

This was not an IT project. It was a productivity exercise. The kind of thing that every business owner says they will get around to but never does, because clicking unsubscribe links one at a time is not a good use of anyone’s afternoon.

The point is not that you should run Graph API scripts yourself. The point is that AI tools like Claude Code turn problems you would normally just live with into problems you can solve in a single sitting. I described what I wanted in plain English, Claude Code wrote the tools, and I made the decisions about what to keep, what to filter, and what to cut.

That pattern applies to dozens of everyday business problems. Data cleanup. Report generation. Process automation. System integration. The gap between “I wish someone would fix this” and “it is fixed” is shrinking fast, because AI can handle the implementation work that used to require hiring a developer or learning to code yourself.

How to Clean Up Your Own Inbox (Three Options)

If this post has you looking at your inbox differently, here are three ways to act on it, from simplest to most thorough:

1. Manual triage (1-2 hours): Search your inbox for “unsubscribe”, sort by sender, and start clicking. Tedious but requires no tools.
2. Use a cleanup tool: Services like Microsoft Viva Insights and third-party tools like Unroll.me can identify newsletter subscriptions. Not as thorough as a full scan, but faster than manual work.
3. Ask your IT provider to run the full triage: If you have managed IT support, your provider can run a scan like this across your mailbox or your entire team’s mailboxes.

If you are a GCIT client and want us to run this for your team, get in touch. It is a quick exercise with a big quality-of-life payoff.

What Would You Use Two Extra Hours For?

The real question is not about email. It is about all the small inefficiencies in your business that you have learned to live with because fixing them seemed harder than tolerating them. AI changes that equation. A 30-minute investment in cleanup that saves you 5 minutes a day gives you back over 20 hours a year. Scale that across your team and it adds up fast.

At GCIT, we use AI tools like Claude Code, Microsoft Power Automate, and custom integrations to solve exactly these kinds of problems for Gold Coast businesses. Sometimes it is inbox cleanup. Sometimes it is automating a process that takes five hours a day. The starting point is always the same: identify the problem, describe it clearly, and let AI handle the heavy lifting.

Elliot Munro is CISO and cofounder of GCIT, a Gold Coast managed IT services provider with ISO 27001, 9001, and 14001 certification. GCIT builds and deploys AI automation solutions for SMBs across South East Queensland.