Ben AI walks through the nine mods he kept after testing more than a hundred, from a sub-agent dashboard to a cache countdown timer.
Posted
6 days ago
Duration
Format
Review
educational
Views
137.2K
2.3K likes
57 · 43
Big Idea
The argument in one line.
Mods let you describe a missing Claude Code feature in plain English and have Claude build it, and the best ones simply listen for events and change the screen without spending any AI usage.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You use Claude Code daily and feel blind during long tasks because you cannot see what sub-agents, tools and files are doing.
You pay for long chats and keep coming back to old ones without knowing why some messages cost far more than others.
You switched from Claude Cowork and miss its side panel showing the to-do list, connectors, skills and files.
You record or share your screen and need client names, emails and dollar amounts hidden automatically.
SKIP IF…
You only use Claude in the browser chat and never open the Claude Code tab or terminal.
You want a deep technical build tutorial, since the video shows what each mod does but not the code.
TL;DR
The full version, fast.
Anthropic launched mods for Claude Code: plugins that listen for signals such as a sent prompt, a tool call or a finished reply, then watch, change or block what happens. You build one by describing it to Claude, which uses its plugin authoring skill. Mods either use AI and draw on your plan, or only display things and cost nothing. Ben demos nine: a brain dump cleaner, a sub-agent dashboard, a cache countdown with keep-warm, a handoff to a fresh chat, a rule saver, a chat inventory panel, a screen share privacy mode, an edit replay, and a switchboard to toggle them all.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Anthropic launches mods; Ben promises nine tested mods and a free pack.
01:04 – 03:18
02 · Mod 1: Brain Dump
Defines how mods listen for signals and shows a mod that turns messy voice dumps into structured prompts.
03:18 – 04:43
03 · Mod 2: Helper Desk
A taskbar and side panel showing each sub-agent's progress and cost.
04:43 – 06:39
04 · Mod 3: Cache Clock
A countdown to cache expiry with a keep-warm option to avoid cold-restart costs.
06:39 – 08:54
05 · Mod 4: Handoff
Moves work to a fresh chat with a detailed handoff prompt before context rot sets in.
08:54 – 10:00
06 · Mod 5: Reflect
Spots hard rules in prompts and asks where to save them.
10:00 – 11:02
07 · Mod 6: Chat Inventory
A Cowork-style side panel of tools, connectors, skills and files.
11:02 – 11:58
08 · Mod 7: Screen Share Mode
Grays out private information while screen sharing or recording.
11:58 – 12:47
09 · Mod 8: Replay Theater
Replays every edit Claude made to a file; works on files only.
12:47 – 14:06
10 · Mod 9: Switchboard and outro
Lists every running mod with on/off toggles, then the free pack and course pitch.
Atomic Insights
Lines worth screenshotting.
A mod is a plugin that listens for events in Claude Code, like a sent prompt or a finished reply, and can then watch, change or block what happens.
You build a mod by describing it to Claude, which uses its built-in plugin authoring skill, and you fix it the same way you fix a skill.
Mods fall into two types: ones that call an AI model and spend plan usage, and ones that only track or display things and cost nothing.
A warm cache reads your chat at a fraction of normal price, so a returning message on a long Opus chat can cost about 4 cents warm versus 1.6 dollars cold.
The cache timer is about one hour on a Claude subscription and only five minutes with an API key, and Claude does not tell you when it runs out.
Sending a small hidden message before the cache goes cold costs a few cents and is far cheaper than one cold restart.
Compacting a chat summarizes older parts, and every summary loses some detail, so answers keep degrading as a chat grows.
A handoff prompt built from the full chat, files, skills and links carries far more context than the short summary Claude writes when it compacts.
A handoff prompt can also be given to a teammate so they pick the task up exactly where you stopped.
Claude auto memory saves what it thinks is relevant, so rules you actually care about often never get saved unless something catches them.
Newer models spin up many sub-agents per task, which is faster but leaves you with no view of each one's progress or cost.
Mods that only redraw the screen from events, like a taskbar or side panel, use no AI and cost nothing to run.
A privacy mode that grays out names, emails, phone numbers and dollar amounts lets you record real chats instead of inventing fake clients.
Once you install many mods you lose track of what is running, the same thing that happened with skills.
Takeaway
Mods make Claude Code visible and cheaper
WHAT TO LEARN
A mod listens for events in Claude Code and can display, change or block what happens, so most of the best ones add visibility and cost control without spending any AI usage.
02Mod 1: Brain Dump
A mod reacts to signals like a sent prompt, a tool call, a sub-agent start or a finished reply, and you can build one just by describing it to Claude.
A brain dump mod can hold back a long messy prompt, ask clarifying questions, and rewrite it into a structured prompt before Claude sees it.
Mods that use AI draw on your plan like a normal message, while mods that only display or track things cost nothing to run.
03Mod 2: Helper Desk
A sub-agent dashboard shows a progress bar, name and cost for each helper, so you know which one is slowing a job down.
A taskbar above the prompt box fills as to-dos are ticked off, so you see how far along the whole job is.
04Mod 3: Cache Clock
Coming back to an old chat after the cache goes cold can multiply a message's cost many times over, so a countdown and a keep-warm ping save real money.
The cache lasts about an hour on a subscription but only five minutes on an API key, and a keep-warm message costs only a few cents.
05Mod 4: Handoff
Long chats degrade as the context fills and compacting loses detail, so moving to a fresh chat with a full handoff prompt keeps both quality and context.
A handoff can trigger at 250,000 tokens, when the cache goes cold, or on demand, and the same prompt can brief a teammate.
06Mod 5: Reflect
Hard rules you state in chat, like never use em dashes, can be caught and saved to built-in memory, Claude.md or a second brain instead of being forgotten.
07Mod 6: Chat Inventory
A chat inventory panel lists connectors, skills, files and to-dos in use, which removes the black box feel of longer tasks.
08Mod 7: Screen Share Mode
A screen share mode grays out names, emails, phone numbers and dollar amounts so you can record or present real chats safely.
09Mod 8: Replay Theater
A replay command plays back every edit Claude made to a file so you can see what changed and return to an earlier version.
10Mod 9: Switchboard
A switchboard lists every running mod so you can toggle each one or hide extra panels, which matters once you install many.
Glossary
Terms worth knowing.
Mod
A plugin for Claude Code that changes how the app looks or behaves in the desktop app or terminal by reacting to events.
Prompt cache
A saved copy of your chat that Anthropic keeps after each reply so Claude can reread it at a much lower price. It expires after a limited time.
Cold cache
A chat whose saved copy has expired, so Claude must reread and store the whole chat again at full price.
Context window
The amount of conversation Claude can hold in short-term memory for a single chat.
Context rot
The drop in answer quality as a chat fills the context window and Claude starts missing details and mixing things up.
Compacting
Claude summarizing older parts of a chat to free up space, which loses some detail.
Sub-agent
A helper instance Claude spins up to research or work on part of a task in parallel while the main chat stays clean.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
Anthropic just launched mods for Cloud Code and I think it's one of the biggest updates they've shipped in a while. With mods, Cloud can now build features for itself. Any feature you wish Cloud had, you just describe it and it builds it right into Cloud Code.
So in this video, I'll explain what a mod is and how easy it is to build one. I'll show you the nine best mods to use after analyzing, building and trying out more than a hundred different ones over the last weeks. And I'll give away all of the mods for free so you can install them for yourself.
Now, if you don't know me yet, I'm Ben, I'm a three -time founder and now run an AI agency and one of the largest AI communities for professionals and business owners where I teach how to use AI to actually automate work or build an AI business. So what is a mod?
A mod is basically a plugin that changes how cloud code looks and behaves, both in the desktop app or in the terminal. Now you can essentially use or build mods that customize how the interface or harness itself looks or behaves, which is really powerful because it can drastically improve the way you work with AI. And it's really a reason to consider switching to the Cloud Code tab if you're still used to working in Cloud Cowork.
The easiest way to explain a mod is by giving you an example of the first mod that I built, which already really improves the way I work with AI, the brain dump mod. Now, many people, including me, do a lot of long voice brain dumps when prompting Cloud. But these brain dumps often get messy.
Voice transcriptions often contain errors. They jump around a lot. They're, of course, very unstructured and sometimes even contradict themselves.
So I built a mod that automatically turns the... messy brain dumps into a clear structured prompt before cloud even processes it and the way a mod works is very simple every time something happens in cloud code it sends out a signal for example when you send a prompt when cloud is about to use a tool when a sub agent starts or when cloud finishes a reply A mod just listens for one of those signals and can then do something when that signal fires.
So my brain dump mod listens for the moment I send a prompt. It then automatically checks if what I just sent is a long, massive brain dump. And if it is, it blocks it.
So the prompt is held back before Claude gets it. And then in this case, it uses another AI model in the background to ask me a few clarifying questions about my brain dump and then turns that into a clear prompt according to Entropic's best practice prompting framework covering that. task, the why, the guardrails and what done looks like.
Cloud can then process that clear prompt instead of my messy brain dump. So that's really what a mod is. You make it listen for a signal and it can then watch, change or block what happens.
And with those few building blocks, you can build almost any feature you want. And making mods yourself is really easy. You just describe what you want and Cloud builds it.
In this case, I said something like build me a mod that catches my long voice brain dumps, ask me a few clarifying questions and then rewrite the prompt into my prompting framework. Cloud then uses its built -in skill called plugin. authoring to build the mod for you.
And if something doesn't work the way you want, you just tell Cloud what to change, just like with a skill. Now, finally, there are basically two types of mods. Ones that use AI, like this one, which uses your plan's usage, just like a normal message, and ones that only track or show things on screen, which costs nothing to run.
Now, all of the mods that I'll show in this video will be in a zip file in the free resources in the first link in the description below, which you can just throw into a Cloud chat and say, install these mods for me, and Cloud will then add them. your own account now the second model is one of my favorites and it's called helper desk in short it shows you exactly what your sub agents are doing while cloud works if you've been using newer models like opus 5 .5 you know they use a lot of sub agents for example every time i ask cloud to help me ideate on a new video it spins up a bunch of sub agents to do the research now this is generally good because cloud can analyze more things faster and keep the main chat clean but the problem with this is there's no real way to see what each sub agent is doing how much it's spending or how far along it is.
So you just sit there not knowing if it will take another five, 10 or 15 minutes. So this mod fixes that in two ways. First, it will give you a purple task bar right above your prompt box that fills up every time a to -do gets ticked off.
So you see how far along the whole job is and which step it's on right now. And second, you get a side panel where every sub agent gets its own little cloth crap with its own color, its own progress bar and what it's costing you. So you can see exactly which one is slowing things down and how they add up.
to the total cost of your chat. Now, this mod just watches every time Cloud starts a sub -agent or ticks off a to -do and then changes what Cloud Code shows on screen by drawing the taskbar and the side panel. So it doesn't use any AI to do that.
So it costs nothing to run. And because it's a mod, you can ask Cloud to change how it looks, for example, to give each sub -agent a name that matches its role. Now, credit to Johnny Savvy's progress mod for this one, which I used as inspiration, but I rebuilt this mod for my own workflow.
Now, the next mod has made a real difference to my token. spend and it's called cash clock. In short it tells you exactly when an old chat is about to get expensive and can keep it cheap for you.
Now if you use cloud a lot you probably have a few long chats you keep coming back to for example after lunch or the next morning but what most people don't know is that coming back to an old chat can cost you far more than you think and that's because of something called the cash. Every time you send a message cloud rereads your entire chat and to make that cheaper Entropic saves a copy of your chat on its own servers after every reply.
which is called the cache or keeping your chat warm. And while the chat is warm, Cloud just reads that saved copy for a fraction of the normal price. On Opus 5 .5, only a 20th of that normal price.
But that saved copy only lives there for a limited amount of time. And once that timer runs out, the chat or cache goes cold, which means that saved copy is gone and Cloud has to reread and save the whole chat again at full price. And on a long chat that can make the same message up to 40...
times more expensive for example on a long opus 5 .5 chat that's about four cents warm versus 1 .6 dollars cold every single time you come back and that timer is actually very short one hour on a cloth subscription and only five minutes with an api key and the problem is you have no idea when that timer actually runs out so the cash clock mod listens for the moment cloth finishes its reply and then shows a countdown until your chat goes cold using the timer straight from entropic's own reply so it's not a guess and if you know you'll come back to the chat later maybe after lunch you just click keep warm or type slash keep warm then right before the cash goes cold it sends a small hidden message that keeps the entire chat warm so the timer starts again that costs only a few cents each time which is of course far less than one cold restart and it stops by itself when you send your next message or after two hours credits to Karan Bensal for this one whose cash tax mod inspired this one now the next mod is one that I built myself and it works really well to get
with cache clock. It's called handoff and in short it moves your work to a fresh chat at the right moment without losing any context from your original chat. Now if you work with cloud on bigger projects you know chats can get really long and the longer a chat gets the worse the answers get.
That's because cloud can only hold so much in its short -term memory for a single chat which is also called the context window. So the fuller it gets the harder it is for cloud to keep track of everything so it starts missing details and mixing things up which is also called context rot. Now, Cloud tries to fix this, of course, by compacting the chat, which basically means it summarizes the older parts to free up space in the chat.
But a summary always loses some details, so the answers still get worse over time. And if the cache has gone cold on top of that, every new message will cost you far more too. So at some point, you do want to start a fresh chat without losing all of the context and progress you've built up, of course.
But it's hard to know when to actually do that. Now, this mod fixes both these problems and it can fire in... three different moments first when your chat hits 250 000 tokens which is sort of the cutoff point for cloth losing a little bit of its performance second when the cash clock sees the cash go cold and third you can just type in slash handoff to trigger it it then reads the whole chat all the files cloud used or created in the chat the skills the tools and the links it used and then writes a well -structured handoff prompt for the new chat that's much more detailed than the short summary cloud makes when it compacts a chat And you then get a simple button to start a fresh chat with that prompt.
So the new chat knows exactly where you left off with all of the context, the progress in the next steps. And because it's just a prompt, you can also use it to hand a task over to someone else on your team so they can pick up exactly where you stopped. But before I show you the next one very quickly, if you enjoy this type of content, you'll really enjoy my AI accelerator where you get access to my 30 day cloud course that takes you from beginner to pro in less than one hour a day, a full course on growing an AI business.
based on everything me and my team have learned over the last three years running an AI agency. You also get access to unlimited one -on -one live AI consulting with AI experts and a big library of skills, templates, and resources across all business departments. So definitely check it out in the link in the description below.
Now, the fifth mod is a really good way to make Cloud actually remember your roles and preferences, and it's called Reflect. In short, it spots when you give Cloud a rule and asks you where you want to save that rule. Now, if you use Cloud Code, you might know it has auto memory.
So it saves some things it thinks are relevant in Cloud's built -in memory. Or if you have a cloud .md or an agents .md and maybe a second brain, it can save it in there too. Now, that's generally really good, but we don't really control what it saves and what it forgets.
And often it just doesn't save things or rules that actually would be great to have in your memory. So this mod fixes that by listening to every... prompt you send in every chat and checking if you just gave Claude a hard role.
For example, something like never use em dashes or keep LinkedIn hooks under 12 words. And if it spots one of those hard rules, you automatically get a pop -up asking if you want to save it to your built -in memory, your Claude .md or your second brain, if you work with one. And if you want your role saved somewhere specific, for example, in a skill, you can just ask Claude to change the mod to however you want it to work.
Now credit to Byram Anokov's Claude Reflect skill. that inspired me to build this mod. Now the next mod is the one that really helped me move from co -work to cloud code and it's called chat inventory.
Now the reason I did many of my tasks in co -work instead of code until recently is because of its UI. If you've used co -work you know it's a side panel where you see the to -do list, the connectors and skills cloud is using and the files it's reading and creating. It's one of the things I like most about co -work because you always know what cloud is actually doing and the cloud code tab just doesn't show you that.
So especially on longer tasks it feels a little bit like a black box so this mod that i built listens for every tool call connector call and skill that cloud uses during a chat and keeps track of them in a side panel called in this chat so just like in co -work you see the to -do list with the goal at the top every connector cloud is calling with its logo and how many times it's called the skills it's using and the files it's reading and creating for example when cloud builds a client report from your crm your email and your drive you can see exactly where it pulled everything from and because it only keeps track of what's happening and doesn't use any AI, it costs nothing to run.
Now, the seventh mod is really useful for me and it's called a screen share mode. In short, it hides private information in your cloud code chats whenever someone else can see your screen. Now, think about how often your cloud chats can end up on someone else's screen.
Maybe you share your screen in a team meeting, on a sales call, or during a training, or maybe you record videos for social media like me, or quick looms to show a teammate how you did something. And the moment you do, everything in that chat is out there. client names, email addresses, deal values, even your own personal details.
And most of the time you only notice when it's already too late. So when you turn this mod on by just typing slash screen share before you share your screen, it changes what cloud code shows on screen. So every name, email, phone number, and dollar amount turns into a gray bar.
Now for me, of course, this one's perfect because I record a lot of my own chats for these videos. And before I had to make up fake client names. Now I just turn on screen share mode and use my real chats.
Now the next mod comes straight from Anthropic and it's called Replay Theater. In short, it replays every edit Clod made to your file so you can see exactly what changed. Now if you do a lot of copywriting or design work with Clod, you know how many edits a single draft can go through.
For example, you ask Clod to type in your newsletter draft, it makes a dozen changes in one go and now you have no idea what it caught or rewrote. And until now, the only way to find out was to compare the versions yourself or ask Clod to list every change. Now this mod...
keeps track of every change Clod makes to your file. After Clod's reply you just type slash replay and it basically plays back all those changes one after the other so you see exactly what Clod changed and you can go back to an earlier version you like better. Now keep in mind it only works on files so your draft needs to be a document Clod edits not just text in a chat.
Now the last mod is another one that I built myself and it becomes really important once you get into mods. It's called Switchboard and in short it lets you see and switch off every mod you're running all in one place now trust me once you start you'll end up building installing and testing a lot of mods and the same thing will happen that happened with skills for many people people installed hundreds of them and i had no idea which ones were actually on their account which ones they're using etc and right now there's no quick way in the desktop app to see every mod that's running or to switch one off in the middle of a chat so this mod adds one simple command when i type slash mods it opens a switchboard with every mod that's running right now.
I can turn each one on or off and hide the extra panels some mods add to my screen. So if a mod gets in the way during a task, I just switch it off and every new mod I add shows up in this list by itself. So I never lose track of what's running.
Now again, you can download all of the mods for free in the first link in description below. And if you want to get access to my full 30 -day cloud course that takes you from beginner to pro in less than one hour a day, unlimited one -on -one AI consulting, a full AI business course and a community with professionals and business owners actively.
using AI you can check out my AI accelerator in a link in the description. As always thank you so much for watching and if you enjoyed this video I highly appreciate a like and a subscribe and if you want to learn more about cloud code you can check out the video here.
The Hook
The bait, then the rug-pull.
Anthropic just let Claude build features for itself, and Ben AI spent weeks testing more than a hundred of them. This is the short list he kept, nine mods that make Claude Code visible, cheaper and easier to live in.
Frameworks
Named ideas worth stealing.
03:10model
Two types of mods
Uses AI (spends plan usage)
Display or track only (costs nothing)
Mods either call a model and draw on your plan, or only watch events and redraw the screen at no cost.
Steal fordeciding which mods are safe to leave on all day
CTA Breakdown
How they asked for the click.
VERBAL ASK
13:33product
“If you want access to my full 30-day Claude course, unlimited one-on-one AI consulting, a full AI business course and a community, check out my AI accelerator.”
Soft pitch repeated mid-video at 8:09 and again at the close, alongside a free mods pack link as the lead magnet.
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
A rapid tour of twelve CLIs, plugins, and connectors that extend Claude Code and Claude Cowork past their default integrations, from a Gmail connector that can't send mail to a scraper that can.
A walkthrough of the rebuilt Claude Code desktop app: new layout cards, projects with nested sessions, side chats, split view, the docs and design tools, and four slash commands worth knowing.
A YouTuber pulls Anthropic's own internal prompting guides for Opus 5 and Fable 5, cross-references them against Claude Code creator Boris Cherny's Y Combinator talk, and distills them into seven concrete habits.
A 43-minute numbered walkthrough of all 34 Claude Cowork concepts across memory, automation, connectors, and team rollout, framed as a business operating system not a chatbot.
Why showing Opus 5.5 a finished example now beats writing out the steps, proven by rebuilding a newsletter skill from six best and six worst performing emails.
00:16
00:30
00:33
00:48
00:58
01:19
01:29
01:40
01:50
02:00
02:11
02:21
02:31
02:41
02:52
03:02
03:12
03:33
03:44
03:55
04:05
04:16
04:27
04:37
04:58
05:09
05:19
05:30
05:41
05:51
06:02
06:12
06:23
06:33
06:44
07:04
07:15
07:25
07:36
07:46
07:56
08:07
08:17
08:28
08:42
08:46
09:10
09:21
09:32
09:43
09:54
10:05
10:15
10:36
10:46
10:56
11:07
11:18
11:41
11:52
12:02
12:12
12:22
12:42
12:52
13:03
13:26
13:37
13:49
14:00
21:41
13:40
13:15
43:17
12:30
09:23