Claude Projects vs Claude Code for personal knowledge

Both of these are built by Anthropic, both are called "Claude," and they solve genuinely different problems. Projects gives you a persistent workspace inside claude.ai. Code gives you an assistant that works directly on files sitting on your own computer. The right one depends on where you want your knowledge to actually live.

Claude Projects is a feature inside claude.ai: you create a project, upload files or paste text as its reference knowledge, set custom instructions for how Claude should behave in that specific context, and every chat inside the project can draw on that shared material. It is a clean way to keep a body of reference material and a consistent set of instructions attached to one ongoing thread of conversations, without re-uploading files or re-explaining context every time.

What Projects gets right

For a defined body of reference material, research for a specific piece of writing, a client's onboarding documents, a set of specs, Projects is genuinely convenient. You upload once, the material stays attached, and every conversation in that project can pull from it. It is a better experience than a single long chat thread once the material gets substantial.

Where Projects stops

The knowledge lives inside claude.ai's project, not as files you can browse, edit, and reorganize outside the app. You can download the files you originally uploaded, but the project itself, its structure, its accumulated context, is not something you can open in a text editor or move to a different tool. That is a real constraint if durability and portability matter to you, the same question that runs through who owns your notion data for every hosted workspace, not just Notion's.

What Claude Code is instead

Claude Code is a different product: a command-line agent that works directly on your own files, wherever they live, in a repository or a plain folder. You talk to it, and it reads, writes, and edits the actual files, runs commands, and reports back. It was built for software projects, where the files are code. Nothing about the underlying pattern, an assistant that can act on plain files rather than just answer questions about uploaded copies of them, is specific to code. That is the whole premise behind building a second brain with AI using the same kind of tool on a folder of your own notes instead of a codebase.

Which one for personal knowledge

If what you want is a place to hold reference material for a specific piece of work and chat about it, Projects does that job well and there is no reason to reach for anything heavier. If what you want is a personal knowledge system that an assistant actively files into, updates, and reconciles over years, plain files on your own machine, worked on through something like Claude Code, keep you independent of any single product's roadmap. That distinction matters more the longer the system is meant to last, which is also why NotebookLM, built around uploaded sources rather than files you own, runs into the same ceiling as Projects for this specific use case.

Nothing stops you from using both, in fact a lot of people do. Projects for a bounded piece of work with a defined set of reference material, and a Claude Code setup pointed at your own ongoing hub of files for the system you expect to still be using in five years. The two are not competitors so much as tools scoped to different time horizons.