Claude Co-work Unlocks AI Agents with New Projects Feature

Claude Co-work Adds Projects to Organize AI Agent Tasks

Claude Co-work, a powerful tool that lets AI agents perform actions like browsing the web and accessing your computer’s files, has introduced a new feature called ‘Projects.’ This update aims to make using these advanced AI agents much simpler by helping users organize different tasks and contexts. The company says this is a major step towards making personal AI agents easy to use, even when you’re not actively at your computer.

For anyone using AI tools, especially those moving beyond simple chatbots, this new project system in Claude Co-work is designed to help manage the complexity that can come with powerful AI agents. It’s a way to keep different AI tasks separate and organized, preventing the confusion that can arise from managing multiple goals and contexts.

Why Projects Matter for AI Context

A key part of using AI effectively is giving it the right information, known as context. Think of context like telling a helpful assistant everything they need to know before they start a job. If you’re working on a personal project, that’s a different set of information than what your AI needs for your main job or for managing personal errands.

The problem has been managing all this different information. Previously, AI interactions could become a jumbled mess. Projects solve this by allowing users to create separate spaces, or ‘projects,’ each with its own specific context. This means you don’t have to keep re-explaining your situation every time you switch tasks.

How Claude Co-work Projects Work

The new Projects feature in Claude Co-work is similar to project organization found in tools like ChatGPT or Claude’s own chat interface. However, Claude Co-work goes further by giving the AI agent direct access to your computer’s folders, remote control of your web browser, and the ability to connect to various other tools.

To start using projects, you simply click a plus button next to the ‘Projects’ tab. You can either create a new project from scratch, giving it a name like ‘AI Education,’ and then add specific files that contain the necessary context. These context files are best kept as markdown (.md) documents, as they are easier for AI to work with and provide more consistent results than PDFs.

Organizing Your Digital Life

Users can set up different projects for different areas of their lives. For instance, one project might be for personal tasks, another for work-related education and community building, and then separate, temporary projects for specific side hustles or new ideas. This framework helps users get started on tasks quickly without needing to gather and re-explain information each time.

When you create a project, Claude Co-work automatically sets up a new folder structure on your device specifically for that project. This is an improvement over older systems where all AI-generated files might have ended up in one large folder, quickly becoming disorganized. These new project folders create clear divisions and dedicated pockets of context.

Inside a Claude Co-work Project

Within a project, you’ll find a place for instructions, scheduled tasks, and specific files. The system also adds custom memories just for that project, helping the AI learn and become smarter about that particular category of tasks without affecting your entire account’s learning. This is similar to how projects worked in other AI chat tools, but now integrated into a more capable agentic application.

The project also links back to the original chat interface, bringing over relevant markdown files and memories dynamically. This ensures that all the context you’ve gathered is available within the project environment.

A Real-World Use Case: Personalized News Briefings

A common and powerful use case for these AI agents is creating personalized news briefings. By setting up a project with personal information and memories, Claude Co-work can generate a custom news summary without needing explicit instructions each time.

For example, a user can prompt the AI to create a personalized news brief. Because the project already contains context about the user’s interests, the AI uses its web search capabilities and browser control to find the latest generative AI news, major global events, and stock market changes related to AI. It then compiles this information into a summary tailored to what the user cares about.

Scheduling Your AI Tasks

Once a task like a news briefing is successfully generated, it can be scheduled. Users can set the AI to run this task automatically, for example, every weekday at 8 a.m. This creates a scheduled task that improves over time. When you review the scheduled task, it shows a more detailed prompt that the AI has built itself, which you can then refine to further improve future outputs.

This includes adding preferred sources that the AI will scan every time the task runs. The developers highlight how user-friendly these agentic applications are becoming, suggesting that most users should try them out.

Availability and Pricing

To use Claude Co-work and its advanced features, users need a Claude Plus subscription, which costs around $20 per month. The company is also offering referral codes that can provide new users with a free week of access, allowing them to try out the workflow before committing to a subscription.

Source: New Claude Cowork Projects Explained in 9 Minutes (YouTube)