Claude CoWork just got a major upgrade, and honestly, this changes the game if you care about AI automation, mobile workflows, and getting real work done without babysitting every step. The biggest shift is simple: Claude CoWork now works across web and mobile, which means you can start tasks from your computer, approve decisions from your phone, and keep work moving in the background even when your laptop is closed.
On top of that, Claude’s interface has been cleaned up, CoWork is easier to access, artifacts are becoming more useful, and the addition of Sonnet 5 and Fable 5 gives you two seriously powerful options depending on what you’re trying to do. If you have been looking for a better way to automate research, content planning, presentations, and repeatable workflows, this is where things start to get very interesting.
If you want the official Claude article on use cases, it is worth reviewing here: How people are using Claude CoWork. For broader AI context and model comparisons, authoritative references like Anthropic can also help.
Claude CoWork now feels built for actual daily use
One of the first things you notice is the redesign. Claude now presents a simpler navigation experience with Home and Code, and CoWork is integrated in a way that feels much cleaner than before. It is a small UI change on the surface, but in practice it makes the product feel more direct.
That matters because CoWork is not a novelty feature. It is quickly becoming one of the best ways to automate useful tasks inside Claude. If your goal is to have AI do more than answer a chat prompt, CoWork is where the real leverage starts to show up.
There is also a temporary boost in usage available, with 2x usage until August 5. If you have access to that, use it. This is exactly the kind of window where it makes sense to test workflows aggressively, push the limits a bit, and figure out what should become part of your normal operating system.
Why Claude CoWork is more powerful than a normal chat
The difference between chatting with Claude and using CoWork is that CoWork is built to take action over time. Inside a CoWork chat, you can choose a specific project context, decide whether Claude should ask for manual approval on each action, or allow it to skip approvals so it can move faster.
If your goal is true automation, turning on skip approvals is a big deal. That removes friction and lets Claude execute a workflow with much less interruption. If you prefer more control, manual approval is still there. The important part is that you now have a much better balance between autonomy and oversight.
A smart way to get started is to use Claude’s own guidance. Claude published a use case article showing how people are using CoWork. A practical move is to drop that article into Claude and ask something like:
- How should I be using CoWork based on everything you know about me?
- Which workflows from this article fit my job or business best?
- What should I prioritize first, and what should I ignore for now?
- Can you turn those recommendations into exact prompts for me?
This works especially well because Claude can combine the article’s ideas with memory and project context. Instead of generic advice, you get a recommendation tailored to your workflow, including what to set up first, what to deprioritize, and what prompts to use.
Using MCP servers with CoWork is where things get really interesting
One of my favorite ways to use Claude CoWork is with MCP servers. When you combine CoWork with outside tools, Claude becomes much more than a writing assistant. It starts behaving more like an operator.
A great example is using the vidIQ MCP server to research YouTube trends. You can ask Claude to find trending content around topics like Claude CoWork, Claude Code, or AI tools in general. Instead of manually opening tabs, checking search results, and compiling notes, Claude can do the legwork for you.
That matters for a few reasons:
- It saves time on repetitive research
- It can gather and organize data quickly
- It can produce outputs that fit your existing workflow
- It can build directly toward the next step, such as outlining or scripting
If Claude already understands how you structure your content, the workflow gets even stronger. Research can feed directly into a script, a brief, or a set of talking points. Instead of stopping at information collection, CoWork can move toward something usable.
That is a big shift from older AI workflows where you had to manually chain everything together yourself.
Mobile access changes Claude CoWork completely
The headline update is mobile. Claude CoWork is now accessible from your phone, and this is not just a nice add-on. It fundamentally changes how useful the product is.
Before this update, if Claude was running a task and needed your input, you often had to return to your computer to keep things moving. That broke the flow. It meant automation still depended on you being physically tied to a machine.
Now that barrier is gone.
With the latest mobile version, you can:
- Start a CoWork task from your phone
- Check progress while away from your desk
- Receive updates directly on mobile
- Respond when Claude needs input
- Review finished outputs from anywhere
This means your work can genuinely follow you. You can begin a task at your desk, step away, and still stay in the loop. If Claude reaches a decision point that requires approval, it can notify you on your phone. You respond there, and the workflow continues.
That sounds simple, but it removes one of the biggest bottlenecks in agentic AI systems: waiting on the human.
Claude can now keep working in the background
Another major shift is that work continues even when your device is not actively open. You can close your laptop, leave for a meeting, and Claude can keep going.
This is huge for scheduled automation. In the past, a lot of AI workflows still felt semi-manual because your device needed to stay involved. Now you can schedule tasks and let them run without keeping a machine online in the same way.
The practical implication is obvious. You should be scheduling work for the hours when you are not actively working.
If you work a standard 9 to 5, use the off-hours. Have Claude:
- prepare briefs
- find opportunities
- collect research
- organize raw information
- surface items that need review
- flag decisions that need your input
Then when you wake up, instead of starting from zero, you begin with a queue of completed or nearly completed work. That is a much better use of AI than just asking one-off questions during the day.
It is also a much better use of your attention. You spend your energy reviewing, deciding, and improving, rather than initiating every tiny step manually.
What the mobile CoWork experience actually unlocks
On mobile, Claude can show you categories like:
- Ready for review
- Needs input
- Active work
That makes it easier to manage AI work in the same way you might manage email, tasks, or project updates. You are no longer forced into a desktop-only model for oversight.
And because notifications can surface moments when Claude needs a choice from you, the automation becomes more resilient. Instead of stalling until you are back at your computer, the process keeps moving as long as you respond.
For anyone trying to build a real AI workflow, this is one of the most important updates Claude has shipped.
Which Claude models you should actually use
Claude now shows several model options, but there are really two worth focusing on here: Sonnet 5 and Fable 5.
Sonnet 5
If you need speed, coding help, agentic tasks, or solid performance without burning through usage as quickly, Sonnet 5 is the practical default. It offers extremely strong output while staying fast and relatively efficient.
The recommendation here is straightforward: if you do not need maximum experimental power for a specific task, use Sonnet 5.
Why Sonnet 5 stands out:
- Fast responses
- Strong coding capability
- Great for agentic workflows
- Less punishing on usage limits
- Affordable if you are using the API
Fable 5
Fable 5 is the model to use when you want something that feels almost ridiculous in capability, especially for creative generation and polished outputs. The catch is timing. It is available only until July 12 before shifting to a usage-based model.
So if you have access to Fable 5 right now, the advice is simple: get as much work done with it as possible while you can.
This is especially true for design-heavy or presentation-heavy tasks where quality matters.
Artifacts inside Claude Code make outputs much more shareable
Another useful upgrade is that you can now see artifacts inside Code. That opens up more possibilities for creating things like:
- dashboards
- live previews
- private share links for teams
This matters because it moves Claude outputs closer to something operational. Instead of producing text and leaving you to reconstruct the result elsewhere, Claude can generate things you can inspect, adapt, and share more directly.
If you are collaborating with a team, even lightweight sharing features can make a big difference. A private link is much cleaner than copying and pasting fragmented work into multiple tools.
The design experience has been upgraded too
Claude’s design workflow has also been improved, and this is where Fable 5 really shows off. You can use Claude to create:
- website prototypes
- PowerPoint style presentations
- documents
- wireframes
- animations
There is also a guest pass system that gives you a few opportunities to invite others and potentially earn usage credits, which is a nice extra if you are getting people into the ecosystem.
But the real story is output quality.
A real example: generating a stock pitch deck with Fable 5
One of the best demonstrations of Fable 5 is using it to generate a polished presentation from a relatively simple prompt.
The example workflow here was a pitch deck comparing the bull and bear case for buying Meta or Microsoft stock at current prices. Not financial advice, obviously, but a perfect test of whether Claude can research, structure, and design something substantial.
The process looked like this:
- Choose Fable 5
- Select a slides or presentation format
- Provide a prompt describing the deck
- Answer a few setup questions about audience, depth, style, and tone
- Let Claude handle the research, layout, and presentation build
Claude asks follow-up questions that actually improve the result. For example:
- How long should the presentation be?
- Who is it for?
- How deep should the numbers go?
- Should the deck take a position?
- What visual style should it use?
- Should it include charts?
- What tone should the copy have?
That kind of guided setup is useful because it helps translate a rough idea into a more production-ready brief. Then Claude takes those inputs, performs the research, and creates the presentation.
The key takeaway is not just that it makes slides. It is that the output can look extremely polished in a very short amount of time, while also doing the research work that would normally slow the whole process down.
And when it is done, you can still edit it, annotate it, adapt it, and share it. So the workflow is not locked. Claude gets you most of the way there, then you take over for refinement if needed.
How to use these updates strategically
If you want to get the most out of these Claude CoWork upgrades, here is the approach I would recommend:
- Use Sonnet 5 as your default model for fast, capable, lower-cost work.
- Use Fable 5 aggressively while available for decks, prototypes, and high-polish creative tasks.
- Turn on skip approvals when you want real automation and trust the workflow.
- Schedule overnight tasks so you wake up to ready-to-review outputs.
- Use mobile for approvals and check-ins so workflows do not stall.
- Connect MCP tools if you want Claude to gather real-world information and act on it.
- Ask Claude to recommend your best CoWork use cases based on your role, projects, and habits.
That combination turns Claude from a chatbot into a work engine.
Suggested media and on-page SEO additions
To make this article more useful and search-friendly, I would include the following supporting media:
- Screenshot of Claude mobile CoWork notifications
Alt text: Claude CoWork mobile notification asking for user input during an automated task - Screenshot of CoWork task dashboard
Alt text: Claude CoWork interface showing tasks ready for review and tasks needing input - Screenshot of Fable 5 presentation output
Alt text: AI-generated investment pitch deck created by Claude Fable 5 - Workflow infographic
Alt text: Claude CoWork workflow from desktop setup to mobile approval and final output review
You may also want to add internal links to related articles on topics like Claude Code, AI automation workflows, MCP servers, or best AI tools for content creation.
FAQ
What is the biggest Claude CoWork upgrade?
The biggest upgrade is mobile access combined with background task continuity. You can now start, monitor, and approve Claude CoWork tasks from your phone, which makes 24/7 automation much more practical.
Can Claude CoWork keep running when my laptop is closed?
Yes. Claude can continue working in the background, and scheduled tasks no longer depend on you actively sitting at your computer in the same way as before.
Should I use Sonnet 5 or Fable 5?
Use Sonnet 5 for most everyday work, especially coding, agentic tasks, and fast execution. Use Fable 5 when you want premium creative output, especially for presentations, prototypes, and design-heavy tasks.
What kinds of tasks are best for Claude CoWork?
Claude CoWork is especially strong for research, content planning, briefs, recurring tasks, MCP-connected workflows, and any process where AI can do most of the work before you step in for review.
Can Claude CoWork create presentations automatically?
Yes. With the upgraded design tools and Fable 5, Claude can research a topic, structure a deck, design slides, and produce a highly polished presentation that you can then edit and share.
How should I start using Claude CoWork if I am new to it?
Start by asking Claude how CoWork fits your workflow based on your projects and goals. Then pick one repeatable task, automate it with Sonnet 5, and add mobile approvals so the process keeps moving even when you are away from your desk.
The bottom line
Claude CoWork was already useful, but these upgrades push it into a different category. Mobile access means your workflows are no longer tied to your desk. Background execution means tasks can keep moving while you are doing something else. Sonnet 5 gives you a practical, high-performance default. Fable 5 gives you output quality that can genuinely surprise you.
Put all of that together, and Claude starts to feel less like a tool you occasionally prompt and more like a system you can delegate work to.
If you are serious about AI productivity, now is the time to experiment with scheduled tasks, mobile approvals, MCP-powered research, and presentation generation. There is a lot of leverage here for anyone willing to set up a few smart workflows.
If this sparked ideas for your own automation stack, explore related AI workflow articles on this site, test a few overnight CoWork tasks, and share what ends up saving you the most time.
