CLAUDE LABJP
MCP — The July 28 MCP spec release candidate drops the Mcp-Session-Id header and goes stateless, so remote MCP servers no longer need sticky sessionsAPPS — The same release adds MCP Apps for server-rendered UI and a Tasks extension for long-running workMEMORY — The Python 0.116.0, TypeScript 0.110.0, and Go 1.56.0 SDKs now send agent-memory-2026-07-22 on every memory store callSPILL — Output from agent_toolset and MCP tools past 100K characters now spills to a file in the sandbox, with the model receiving a truncated preview it can expandBG — MCP tool calls running past two minutes move to the background automatically, keeping the session usable; tune it with CLAUDE_CODE_MCP_AUTO_BACKGROUND_MSRESUME — Typing /resume in the agent view opens a picker of past sessions and brings your pick back as a background sessionMCP — The July 28 MCP spec release candidate drops the Mcp-Session-Id header and goes stateless, so remote MCP servers no longer need sticky sessionsAPPS — The same release adds MCP Apps for server-rendered UI and a Tasks extension for long-running workMEMORY — The Python 0.116.0, TypeScript 0.110.0, and Go 1.56.0 SDKs now send agent-memory-2026-07-22 on every memory store callSPILL — Output from agent_toolset and MCP tools past 100K characters now spills to a file in the sandbox, with the model receiving a truncated preview it can expandBG — MCP tool calls running past two minutes move to the background automatically, keeping the session usable; tune it with CLAUDE_CODE_MCP_AUTO_BACKGROUND_MSRESUME — Typing /resume in the agent view opens a picker of past sessions and brings your pick back as a background session
Articles/Cowork
Cowork/2026-03-15Beginner

Claude in Chrome vs. Cowork: Watch It Work, or Let It Run

The real difference between Claude in Chrome and Cowork isn't features — it's how you work with each. First-day pitfalls and safe starter prompts included.

Claude47Cowork33Claude in Chrome14beginner11getting started

Two things open on my screen every morning: a browser, and a working folder. Checks that live inside the screen — AdMob reports, Search Console graphs — happen in the browser. Drafts, exports, and the other files that live on my desk get handled on the folder side.

Claude in Chrome and Cowork map onto those two views almost exactly. When a friend who had just started with Claude asked me which one to install, I noticed that reciting feature lists didn't help at all — what made it click was explaining where each tool works and how you work with it. Here is that explanation, written out properly.

The difference is where they work: in the browser, or in your folders

Claude in Chrome is an extension you add to Chrome. Claude looks at the same page you have open and does the clicking and typing for you, right there. Think of a colleague leaning over your shoulder — eyes on the same screen, hands on the controls.

Cowork is a desktop app. It works inside folders you choose, reading files, organizing them, and producing new documents. This one is closer to a colleague with their own desk: you hand over a task, they decide how to proceed, and you collect the result.

A companion at your screen, and a colleague at their own desk. Nearly every "which one do I need?" question resolves from that single picture.

When Claude in Chrome is the right tool: it borrows your logged-in screen

The extension's quiet advantage is that it works with pages you are already signed in to.

Admin dashboards are usually visible to no one but you. In my case that means AdMob, where I check ad revenue for the apps I build and run on my own, and Search Console, where I watch search traffic. With a report open, I ask "what changed since last month, in three points?" and Claude reads the numbers straight off the screen. No exporting, no pasting data into another tool — in daily use, that is the part I appreciate most.

It is well suited to work like this:

  • Reading the open page and producing summaries or transcriptions
  • Operating the page on your behalf — filling forms, changing settings
  • Walking through several pages in sequence to gather information

One expectation worth setting early: Claude in Chrome operates in front of you. Keep the tab open and watch the work happen. The "leave, come back, find everything done" style of working belongs to the other tool.

When Cowork is the right tool: hand it over and step away

Cowork is built for delegation. Choose a folder, hand over the job, and move on to something else while it runs.

  • Renaming and sorting files in bulk, following a rule you describe
  • Reading several documents side by side and producing a list or summary
  • Writing out reports as Word, Excel, or PDF files
  • Registering recurring work as scheduled tasks that run daily or weekly

I run several tech blogs, and Cowork's scheduled tasks now carry their routine upkeep — organizing article files, recording update logs. The same procedure runs at the same hour every night. The biggest change was not speed; it was that "forgot to do it" quietly disappeared from my week.

Scheduling is Cowork-only — Claude in Chrome has nothing like it. If your week contains fixed, repeating chores, that alone can settle the choice. And for the days when a registered task does not behave the way you expected, I keep the fixes in a separate piece: why scheduled tasks stop and how to recover.

Still unsure? Two questions settle it

Picture one task you want to hand over, then ask yourself:

Does it need eyes on the screen? If you want to check the display as it changes, or have numbers read out of an admin panel, that is Claude in Chrome. Web pages shift state constantly; a tool that works while you watch suits them.

Is it a fixed, repeatable procedure? "Organize these files the same way every week" or "compile this document at month-end" — that is Cowork. The more settled the procedure, the more value there is in being able to walk away while it runs.

"Browser versus files" works fine as a first rule of thumb. After months of using both every day, though, I find the deeper split is in the relationship: watch it work, or let it run.

Three stumbles almost everyone hits on day one

These are, almost word for word, the first three questions I received after recommending the tools to people around me.

1. "Claude in Chrome isn't responding."

Installing the extension is not enough on its own. Check three things in order: you are signed in to claude.ai, the extension is pinned, and the site you want to work on has been granted permission. Closing the tab you asked it to work on also stops the work mid-task.

2. "Cowork can't see my files."

Cowork's working range is exactly the folders you selected — nothing more. If your desktop files seem invisible, check whether that folder was actually added as a workspace. The flip side is reassuring: folders you have not selected stay untouched. That clear boundary matters the first time you let an AI near your files.

3. "It stopped when I closed my laptop."

Both tools work on your computer, not somewhere in the cloud. When your machine sleeps, Claude rests too. Before handing over a long job, it is worth reviewing your power and sleep settings once.

Start with prompts that can't break anything

Until the behavior feels familiar, begin with read-only requests that change nothing.

  • For Claude in Chrome: "Summarize the key points of this page in three bullets."
  • For Cowork: "List everything in this folder, grouped by file type."

Neither request deletes or rewrites anything — a safe first step. Once you have a feel for how each tool moves, graduate to hands-on requests like form filling and file reorganizing. Claude asks for confirmation before consequential actions, but I would still treat the first few runs as trust-building with small jobs.

A companion at your screen, and a colleague with their own desk. With both in place, the range of work you can hand off widens quickly. Pick one repetitive task from today and pass it to the right one — I hope this way of sorting them saves you the hesitation I watched my friend go through.

Share

Thank You for Reading

Claude Lab is ad-free, supported entirely by members like you. We publish practical guides daily with implementation code, benchmarks, and production-ready patterns. If you've found it useful, we'd love to have you on board.

  • Copy-paste ready implementation code
  • New advanced guides published daily
  • $5/mo or $10 for lifetime access
View Membership →

If you found this article helpful, a small tip ($1.50) would mean a lot to us. Your support helps keep this site ad-free and covers server and hosting costs.

Related Articles

Cowork2026-06-04
Where the human still has to press the button — lessons from delegating ad ops to Cowork
From actually delegating repetitive ad operations — AdMob bidding-partner sign-ups and floor tuning — to Cowork, here is the boundary I found between what AI can fully own and the single moves a human still has to make by hand.
Cowork2026-05-29
Three Weeks of Letting Claude in Chrome Tune My AdMob Mediation Priorities
I let Claude in Chrome handle the reordering of AdMob mediation priorities for three weeks. Here is how I set the threshold, the three prompt changes that actually mattered, the numbers, and the parts I deliberately kept under human control.
Cowork2026-05-28
Folding AdMob and Crashlytics into One Morning Check via Cowork Scheduled Tasks — Two-Week Notes
I had been checking AdMob fill rates and Crashlytics surges in two separate dashboards each morning. I folded them into a single Cowork scheduled task. Here are my two-week notes, with the numbers and the friction I ran into.
📚RECOMMENDED BOOKS
Build a Large Language Model (From Scratch)
Sebastian Raschka
LLM Dev
Prompt Engineering for LLMs
Berryman & Ziegler
Prompting
AI Engineering
Chip Huyen
AI Eng
* Contains affiliate links
See all →