Launching today
Claude Cowork Projects
Tasks, context, and files organized in one workspace
428 followers
Tasks, context, and files organized in one workspace
428 followers
Projects in Claude Cowork bring your tasks, files, instructions, and memory into one focused workspace on your desktop. Built for individuals and teams managing complex workflows, it keeps everything local, organized, and reusable so you can work faster with consistent context across tasks.
Really interesting concept.
I noticed some UX friction around project structure and how users understand the difference from standard Claude workflows.
Happy to share a quick teardown with suggestions if useful.
Claude Cowork Projects turns Claude Desktop into a structured workspace for ongoing AI tasks.
Instead of one-off chats, you can organize work into project-based environments with local folders, instructions, memory, and scheduled tasks solving the chaos of fragmented workflows.
What makes it stand out is its local-first, project-scoped approach with persistent context and memory, plus recurring task automation inside each project.
Key features include project-specific instructions, memory, scheduled tasks, and folder-based context.
Perfect for researchers, operators, and teams running repeatable workflows like reporting, analysis, or knowledge work.
P.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends
@rohanrecommends Yeah this is solid. You actually made it make sense without overdoing it. The local-first + project memory angle is the real hook here. That’s what makes it feel useful, not just another AI tool.
Feels like something people would actually use daily, not just test and forget.
One thing I’m curious about though. If a team is using this together, does the context stay organized or does it start getting messy after a while?
The issue I have with Cowork and also with local projects is, that my local file system is a mess. I have several versions of the same file and it's hard for anyone who is not me, to figure out which is the most current version. While in the usual online projects, I work clean and always delete and replace the most current one. It's not just file versions, but also the overall focus that's different. I still try to find a way to fit Cowork into my workflow and figuring out what the advantage is compared to online projects.
The local-first approach with persistent context is the right call — the biggest pain with AI workflows isn't capability, it's continuity. Every session restart wipes your mental model. I've been building in this space and context bleed between tasks is the #1 productivity killer. What I'd love to see: explicit context scoping per task (not just per project), so you can run parallel subtasks without them leaking into each other. Also curious how you handle scheduled task failures — does it surface errors with enough context to recover without starting over?
the context management is the part that matters most. i run a full business off claude code and the biggest friction is keeping project context across sessions. how does this handle large codebases?
Scheduled tasks within project context is the killer feature here. The biggest pain with Claude Desktop was losing context between sessions. Does the memory persist across model updates or does it need retraining?