Claude Can Now Run Your Other Chats: We Tested the Master Chat on 25 Real Chats

You open Claude in the morning and there are twenty-five chats in the sidebar. One of them asked you a question yesterday, and you have no idea which one. The fix: Claude can now run your other chats for you. One pinned chat checks the rest, tells you exactly what's waiting on you, and even starts new chats on its own.

We tested it on a real job, planning a family trip to Japan, across 25 chats. It found every waiting question, opened four new chats by itself, and checked its own answers. Our weekly usage went from 51% to 52% after everything in this post. Here are the 10 updates that matter if you don't code, plus one bonus and what didn't work.

You don't need to code

The desktop app for Claude now has two tabs, Chat and Code, since Cowork was folded into Chat last month. Every update below lives in the Code tab. Don't let the name put you off: you type plain English, and you never see a terminal window.

1. The master chat

In the new app, one chat can see your other chats. Pick one, pin it to the top of the sidebar, and tell it: you're my master chat. That's the whole setup. Then ask one question: what's waiting on me?

We started with five chats, two of which ended with a question for us: the hotel chat (pay more to stay near the train station?) and the visa chat (passport expiry date?). The master chat came back with exactly those two, each with a link straight to that chat, and said the other three were done.

The 25-chat test (and the cost)

Five chats is easy. Here's the catch most people miss: by default, the master chat only looks at your 20 most recent chats. So what happens when the important one is chat number 21?

We piled on twenty more trip questions, then asked a brand-new master chat the same question. It read 32 transcripts, more than the default, and still found both. Unprompted, it also flagged two answers that quoted train prices from memory and told us to check them before buying anything. It's not just finding chats. It's checking them.

And the cost. Before we started, the usage screen said 51% of the week. After all 25 chats and three master checks, it still said 51%.

5 master prompts for everyday jobs

The trip was just one example. A few we'd actually use:

• Moving house: "Open separate chats for the checklist, the utility switch-overs, the packing plan and the change-of-address list. Tell me each morning what's waiting on me."
• A small business: "I run a bakery. Open separate chats for this week's social posts, the supplier order and replies to customer emails. Never send anything without asking me."
• End of day: "Give me a one-line status for every chat you started today, and list what's still waiting on me tomorrow."

All five are in the free playbook.

2. Chats that report back

We asked the master chat to plan the whole trip and split the work up. It opened four new chats on its own: the day-by-day plan, the budget, the packing list, and passports and visas. They sit right under the master chat in the sidebar, and the master keeps a live checklist of all four.

A few minutes later, three were done and the fourth showed an orange label: needs your input. These are real chats, not hidden background helpers. We clicked through and answered two questions on a simple card. It went to the official sources, wrote a checklist with dates, and the master chat posted an update on its own: all four trip chats finished, no visas needed. One honest note: the government website blocked it with a bot check, so it told us to check that page ourselves.

3. Sections

Sections are folders: click the three dots next to a chat and choose Move to section. Or tell the master chat to sort your chats into sections, as we did with "trip plan" and "quick questions". Chats the master started stay grouped with it.

4. The morning check (scheduled)

The master chat only checks when you ask. A scheduled task fixes that. We typed: "Set this up as a scheduled task: every weekday at 8 AM, check my Japan Trip chats and tell me what's waiting on me." It built a read-only task and gave the fine print up front: the app has to be open at 8, and it runs a few minutes later, around 8:12. A test run checked 31 chats and listed what was still waiting.

5. Projects

A Project runs on Anthropic's computers, so it keeps working after you close your laptop. A master chat runs on your own computer, so it can use your files but stops when your computer sleeps. If the job needs your files, use a master chat. If it's a big job to run while you're away, use a Project, and keep private files out of it. Projects are still in beta.

6. Side chats

Asking a question in the main chat mid-job interrupts the work. Instead, press Command+semicolon on a Mac, Control+semicolon on Windows, or type /btw. A side chat opens that can read everything the main chat has done. We asked which waiting chat to answer first. It said the passport, because a renewal takes weeks.

7. Split view

Right-click a chat, choose Open in, then Split view, or drag it into the window. Start with two. Past that, your own head runs out of room before Claude does.

8. Your phone

Turn on the Remote Control switch in a chat's right panel and you can pick it up from Claude on your phone or the web. Also turn on Keep computer awake under Settings, then System. Closing the lid still sleeps the laptop; for that, use a Project.

9. Docs and slides

We asked the itinerary chat for a one-page document and a five-slide version for the family. It did both, and the slides download as a PDF or PowerPoint. Then we asked it to make day six easier. It changed only that day, then pointed out the swap had dropped a temple visit from the trip and that the slides still showed the old day six. One click, and slide four was fixed. It caught the side effect of our own change.

10. Dashboards

The budget chat built a shareable dashboard showing the trip at three price levels. Flights alone are 44% of the mid-range total. With the new AI-powered artifacts setting (under Capabilities) on, Claude added a question box for things like "how much is left for food?", but in our test it wouldn't answer in the app or on the website. Treat this one as early.

Bonus: the undo button

Type /recap for one line on where a chat is up to. Type /rewind for an undo button: we had Claude rewrite the Kyoto weather like a pirate, picked the message before it, and the pirate was gone. There's also Undo rewind.

What didn't work

The master chat only checks when you ask, unless you schedule it. The passport chat couldn't open the government website, and said so. The dashboard question box didn't answer. And chats the master starts are tied to it, so they can't be moved on their own. Total cost of everything in this post: the week went from 51% to 52%.

3 mistakes to avoid

1. Asking questions in the main chat while it's working. Open a side chat so you don't interrupt the job.
2. Trusting prices and dates from memory. Ask it to check the official source, and check the important ones yourself.
3. Letting it tidy up without rules. When you ask it to sort your chats, add one line: don't archive or delete anything.

Your 60-second setup

1. Open the desktop app for Claude and click the Code tab.
2. Pick a folder for the project, even an empty one on your desktop.
3. Start a new chat and pin it to the top.
4. Paste: "You're my master chat. Keep an eye on my other chats and tell me when one needs me."
5. Add the morning check with one more sentence, and turn on Keep computer awake under System if your jobs run long.

Which ones to turn on today

More than three chats going? Pin a master chat. Open Claude every morning? Add the morning check. Share plans? Try docs and dashboards.

Get the Captain Chat Playbook

Every prompt and setting from this test is in one free playbook. Hand the PDF to Claude and say "set this up for me." Comment CAPTAIN on the video, or grab it below.

How we tested: Claude desktop app 3.6556.2, Max plan, Opus 5.5 Medium, October 28, 2026. Five chats with two waiting questions, then 25 chats checked from a brand-new chat. Weekly limit: 51% before, 51% after the 25-chat test, 52% after everything. One small test, not a study. Independent, not sponsored by or affiliated with Anthropic.