I about Claude Code and the way you need to use it to automate processes and be extra environment friendly as an engineer. Nevertheless, Anthropic additionally has a instrument known as , which you’ll entry both via their net interface or their utility.
Cowork usually targets much less technical individuals with a less complicated UI whereas nonetheless preserving plenty of the capabilities that you’ve in Claude code. On this article, I’ll talk about how one can get essentially the most out of Claude Cowork. In fact, that is very helpful to non-technical individuals who don’t wish to work together with a coding agent via a terminal, however I additionally consider Claude Cowork will be helpful for technical customers with options akin to
- A cleaner interface
- Visualizing figures immediately within the app
- Extra deal with interactive UI parts
On this article, I’ll cowl on a excessive stage the best way to get essentially the most out of Claude Cowork with some particular methods that I make the most of.
I’m not sponsored by Anthropic in writing this text; I’m merely a consumer of their product.
Why use Claude Cowork
I feel there are two foremost sides to why you need to use Claude Cowork. One facet is if you happen to’re a non-technical consumer. Should you don’t work a lot with coding brokers, Claude Cowork is a superb introduction. It simplifies the interplay loads, the place you merely work together with Claude via an online interface or Mac/Home windows utility.
Cowork is, in some ways, only a less complicated strategy to entry the highly effective options of Claude Code. You get all the identical capabilities and choices as you get while you work together with Claude Code via a CLI, but it surely’s offered to you in a nicer method, and it makes it straightforward to get an summary of all of the duties you’re engaged on.
Lots of people would possibly assume Claude Cowork is totally ineffective if you happen to’re a technical consumer and are already utilizing Claude Code via your terminal. Nevertheless, I’d argue the other, and I do discover myself utilizing Claude Cowork at instances the place I would like extra deal with the visible expertise, or I’ve separate, much less technical duties that I don’t wish to carry out by way of the terminal.
Learn how to successfully use Claude Cowork
Now, let’s transfer on to how you have to be utilizing Claude Cowork. Should you’re already aware of coding brokers akin to Claude Code, you need to attempt to use Cowork in the very same method. Basically, specializing in isolating duties, offering clear prompts, working in plan mode, and ensuring the agent opinions its personal output.
Nevertheless, if you happen to’re not aware of coding engines from earlier than, I’ll undergo every of those factors in additional element under.
Isolating duties
Claude not too long ago upgraded its strongest mannequin, Opus 4.6, to have a 1 million context window. A 1 million token context window is gigantic, and you may match plenty of context into the mannequin on the identical time. Nevertheless, although the mannequin can’t match 1 million tokens on the identical time, I urge you to nonetheless reduce the token utilization. The rationale for that’s that efficiency degrades the extra tokens the mannequin has in reminiscence. In brief:
The much less tokens you’ve gotten within the context, the higher efficiency shall be.
Now, you naturally need to steadiness this with the truth that not offering sufficient context to the coding brokers will even degrade efficiency. For instance, if you happen to’re not offering info to the coding agent on which instruments can be found or particulars on the best way to full duties, clearly, the mannequin is not going to carry out nicely. The purpose I’m making above is that you need to keep away from pointless info within the context window of the mannequin as a result of that can degrade efficiency, not solely due to the noise the mannequin has to take care of, but in addition just because the extra textual content the mannequin has to take care of, the extra difficult the issue will change into for the mannequin.
An instance of the best way to successfully scale back the dimensions of the context window is to start out a brand new thread or chat each time you begin a brand new process. Merely put, a brand new process ought to by no means be executed in the identical thread as a earlier process that you’ve accomplished.
It’s additionally good follow to cut back the quantity of MCPs and connectors the mannequin has out there, so long as these MCPs and connectors aren’t related to the issue at hand. Anthropic has already handled this downside fairly nicely by loading in such options dynamically solely when related. Nevertheless, merely eradicating pointless MCPs or connectors will get rid of the issue altogether.
Should you look in the proper facet menu of Claude Cowork, you can too see a context tab highlighting every thing that you’ve in context at present. That is proven within the picture above. We will make sure that you take away any context that isn’t obligatory anymore.
Lastly, one other strategy to isolate duties is that you just at all times guarantee you’ve gotten totally different folders for the totally different areas you’re making use of Claude Code to. For instance, in case you have a bunch of shows generated with Claude Cowork, you need to have a separate shows folder that you just at all times open everytime you wish to make a brand new presentation. And in case you have a separate gross sales undertaking happening, you also needs to make a separate folder for that and work in that folder everytime you wish to do gross sales with Claude Cowork.
Clear prompting
Clear prompting is one other crucial factor to remember when working with coding brokers. Should you gave your immediate to a human and the human wasn’t in a position to perceive the immediate and your intent with it, you’re making it onerous for the mannequin, which can once more result in worse outcomes. Not offering specific and clear prompts merely makes it much less seemingly that the agent is ready to implement what you meant.
One other necessary level right here is to make use of plan mode everytime you begin engaged on extra difficult duties. Should you’re merely performing a simple process, you don’t essentially want to make use of plan mode, however as a basic rule of thumb, I urge you to at all times use plan mode everytime you begin a brand new process. This makes the mannequin assume extra about the issue you’ve given it, ensures that it asks clarifying questions in an effort to elaborate on something that’s unclear, and, basically, improves the efficiency of the mannequin.
You must think about plan mode as a instrument that aligns your intent with the agent’s implementation.
Reviewing agent output
Reviewing the agent’s output shouldn’t be an important factor to do if you wish to get essentially the most out of Claude Cowork. One of many nice advantages of Claude Cowork is that it makes it simpler to visualise issues which are onerous to visualise in a terminal the place you utilize Claude Code. For instance, you may ask the mannequin to attract up illustrations or create diagrams. For instance, if you happen to’re engaged on a sophisticated structure, you may ask the mannequin to attract out the structure, and it’ll create a pleasant flowchart for you.
The picture under highlights how one can work with diagrams in Claude Cowork. On this occasion, I requested Claude Cowork to generate a flowchart of a gross sales pipeline. And as you may see, the mannequin creates the flowchart on the right-hand facet, and I can proceed engaged on the flowchart whereas having Claude Cowork on the left-hand facet. This makes it extremely straightforward to work with visuals inside Claude Cowork, which is likely one of the main benefits of working in Cowork reasonably than Claude Code.
Abilities
The final half I wish to cowl is to make use of abilities everytime you’re engaged on totally different duties. Should you’re performing any repeatable duties, akin to producing shows, you need to make it right into a ability. This ability is basically only a immediate that you’ve saved on the best way to carry out a particular process. I, for instance, have a presentation ability that is aware of precisely:
- What sort of theme do I would like in a presentation
- Roughly the quantity of textual content and textual content dimension I would like within the presentation
- The place I usually need my pictures,
- The alignment I would like in my shows
And each different desire that I’ve.
Conclusion
On this article, I’ve gone into extra element on the best way to get essentially the most out of Claude Cowork. Claude Cowork is a instrument similar to Claude Code that may primarily carry out all the identical duties. Nevertheless, it does have some advantages that Claude Code doesn’t have, particularly for much less technical customers. I highlighted how one can get essentially the most out of Claude Cowork by isolating duties, utilizing clear prompting methods, and reviewing brokers’ outputs. Although I’m a technical consumer myself, I do discover myself utilizing Claude Cowork at instances if I wish to visualize outcomes higher or just wish to work on some duties with a clearer interface. I urge you to check out Claude Cowork your self to expertise what it’s like, and you may decide for yourselves whether or not you wish to carry out some duties in Claude Cowork or if you wish to use Claude Cowork for every thing. Nevertheless, basically, I urge you to attempt to use coding engines to finish as many duties as potential, as they’re extremely highly effective fashions which are in a position to full plenty of cognitive work.
👉 My free eBook and Webinar:
🚀 10x Your Engineering with LLMs (Free 3-Day E-mail Course)
📚 Get my free Imaginative and prescient Language Fashions e book
💻 My webinar on Imaginative and prescient Language Fashions
👉 Discover me on socials:
💌 Substack
