← All articles

vs

Claude Code vs Claude Projects: when a chat project stops being enough

By Tim Siegelkow · Published October 25, 2026 · Updated October 25, 2026 · 2558 words

Full disclosure: I built the AI-OS, so I'm biased. It shows up where it genuinely fits; every other tool here is described from its own documentation and pricing pages. Prices and features change.

Here is the short answer. If your work is mostly talking things through with Claude, with a stable set of reference documents, Claude Projects is enough, and it is the simpler choice. If you want Claude to actually do the work, open your files, change them, save new ones and use other tools, you have outgrown Projects and want Claude Code.

Both come with a paid Claude subscription. So this is not a question of which one to buy. It is a question of which one fits the job in front of you. Plenty of people only ever need Projects, and I will say so again further down, because it is true.

Let me rebuild both from the ground up first, because the names make them sound closer than they are.

The one-sentence version of each

Claude Projects is a workspace inside the normal Claude chat. You give it a name, upload documents to its knowledge base, write a few instructions, and then every chat you start inside that project already knows that background. Anthropic describes projects as "self-contained workspaces with their own chat histories and knowledge bases" where you "upload documents, provide context, and have focused chats with Claude" (Claude Help Center).

Claude Code is an agent that works inside a folder on your computer. It reads the files in that folder, edits them, runs commands and connects to your other tools. Anthropic's own description: it "reads your codebase, edits files, runs commands, and integrates with your development tools" and is available "in your terminal, IDE, desktop app, and browser" (Claude Code docs).

An analogy that helps. A Project is a very smart colleague sitting across the desk, with a binder of your documents. You talk, they answer, you copy what they wrote into your own files. Claude Code is the same colleague, but now they have a key to your office. They can open the filing cabinet, rewrite the proposal in place, file the new version and send the result to the right tool. More useful, and also more responsibility.

Do not let the word "code" scare you off. Claude Code works on any files, not just software. I wrote a whole piece on that: Claude Code for non-developers.

Side-by-side comparison

Claude Projects Claude Code
Where it runs Inside Claude chat (web, desktop, mobile) Terminal, IDE, desktop app and browser, working on a folder
Instructions Project instructions you type into the project A CLAUDE.md text file in your folder, loaded every session
Memory Project knowledge base plus the project's chat history CLAUDE.md files plus auto memory, notes Claude writes itself
File access Files you upload; it reads them Reads, edits and creates files in your folder directly
Tools Connectors to your apps from the chat Connectors through MCP, plus running commands
Automation You start each chat Can run prompts on a schedule
Plans All plans, free users up to five projects Pro and Max; not Free
Learning curve Very low Moderate; the first install is the hardest part

Every row is explained, with sources, below.

Where each one runs

Projects live in the Claude app you already use. Nothing to install. You open a project, start a chat, and the project's knowledge and instructions are there.

Claude Code runs on several surfaces. The docs list "the terminal, IDE extensions, a desktop app, and the web" (Claude Code docs). A terminal is the text window on your computer where you type commands instead of clicking. Most non-developers meet it exactly once, to paste the install line, and the docs even link a guide for people who have never opened one.

The real difference is not the window. It is what Claude is pointed at. A Project is pointed at a conversation. Claude Code is pointed at a folder.

Instructions and memory

In a Project, you write project instructions, for example "use a formal tone" or "answer as a tax advisor would", and you upload documents to the knowledge base. Anthropic says Claude uses that knowledge "to better understand the context and background for your individual chats within that project" (Claude Help Center). On paid plans, when the knowledge base gets large, projects switch to retrieval (RAG, meaning Claude looks up the relevant pieces instead of reading everything at once) to "expand capacity by up to 10x" (same source).

Claude Code does the same job with plain text files. Each session starts fresh, and two mechanisms carry knowledge across: "CLAUDE.md files: instructions you write to give Claude persistent context" and "Auto memory: notes Claude writes itself based on your corrections and preferences" (Claude Code memory docs).

That sounds like a small difference. In practice it is the big one. Your Project instructions sit in a settings box. Your CLAUDE.md sits in your folder, next to your work, where you can edit it, copy it, back it up and grow it over months. If you want to write a good one, start with how to write a CLAUDE.md file.

File access: reading versus doing

This is where Projects stop being enough for most people who leave them.

In a Project, Claude reads what you uploaded. When it writes something useful, a proposal, a client email, a content plan, you copy it out by hand. When the source document changes, you upload the new version. The knowledge base is a copy of your work, not your work.

Claude Code works on the real files. The docs put it plainly: it "edits files, runs commands" (Claude Code docs). Ask it to update last month's client report with this month's numbers and it opens the file, changes it and saves it. Ask it to turn ten meeting notes into one summary and it writes the summary into your folder.

If you notice yourself copying and pasting between the chat and your documents every day, that is the signal.

Tools and MCP

Both can reach other apps, but the depth differs.

Claude chat lets you "connect your apps and tools", and that is listed even on the free plan (Claude pricing). So a Project is not cut off from the world.

Claude Code connects through MCP, the Model Context Protocol, which is a standard way to plug a tool into Claude. The MCP docs cover remote servers, local servers, and scopes so a connection can belong to one folder or to your whole machine (Claude Code MCP docs). On top of that, Claude Code can run commands on your computer, which means any tool with a command line becomes reachable too.

For a coach, that might be calendar, email and a notes app. For an agency, a project tracker and a shared drive. The difference is that in Claude Code these tools sit next to your files and your instructions, so one request can read a file, check the calendar and write a draft in one go.

Automation

A Project waits for you. Every chat starts because you started it.

Claude Code can run on its own. The scheduled tasks docs describe cloud tasks "for work that should run reliably without your machine", desktop tasks "when you need access to local files and tools", and a /loop command for repeating a prompt while a session is open (Claude Code scheduled tasks).

That is the line between a helper and a system. "Draft tomorrow's client replies before I wake up" is a Claude Code job, not a Project job.

A word of caution from my own work: do not put anything on a schedule until you have watched it do the job well by hand a few times. Earn the autonomy first.

Plans and price

Projects are available on every plan. Anthropic says they are "available to all users, including those with free Claude accounts" and that "free users can create a maximum of five projects" (Claude Help Center). Sharing a project with colleagues is a Team and Enterprise feature (same source).

Claude Code needs a paid plan. On the pricing page's feature table, Claude Code is marked "No" for Free and "Yes" for Pro, Max 5x and Max 20x (Claude pricing). Pro is listed at $20 a month billed monthly, or $17 a month with the annual discount; Max starts from $100 a month (same source). Anthropic's help center confirms that with Pro or Max you get "both Claude on the web, desktop, and mobile apps and Claude Code in your terminal with one unified subscription" (Claude Help Center).

One trap worth knowing: if your computer has an ANTHROPIC_API_KEY set, Claude Code uses that key instead of your subscription, and you pay API charges (same source). Most non-developers will never have one set, but if a bill appears that you did not expect, check that first.

So if you already pay for Pro, you already have both. The decision costs you nothing but time.

One thing that is changing: the new projects

Anthropic is blurring this line. A new version of projects is in beta, "rolling out in stages, starting with Claude Code". In it, "a project is one conversation" and Claude "breaks the work into parallel threads that run in the cloud", with each thread starting from the project's files, instructions and memory (Claude Help Center). It is available to select Pro and Max subscribers for now, with chat and other plans to follow (same source).

So the gap between "a chat project" and "an agent" is shrinking. Everything in this article describes the version most people have today. If you see "Projects" inside Claude Code already, you are in the beta.

Who should choose Claude Projects

Choose Projects, and stay there happily, if this sounds like you:

Most coaches and consultants who mainly want a sharper writing and thinking partner are well served by two or three good Projects. Set up one per client or per offer, write clear instructions, upload your best examples. That alone beats generic chat by a wide margin.

Who should choose Claude Code

Move to Claude Code when one of these is true:

If you are deciding between Claude Code and a coding editor instead, that is a different question, covered in Claude Code vs Cursor.

How to move from Projects to Claude Code

You do not have to throw anything away. Here is the path I would take.

  1. Make a folder. One folder on your computer for your business, for example my-business. Inside it, a subfolder per area: clients, content, offers.
  2. Bring your project knowledge over. Download or copy the documents you uploaded to your Project into the matching subfolder. They are just files.
  3. Turn your project instructions into a CLAUDE.md. Open a new text file called CLAUDE.md at the top of the folder. Paste your project instructions in. Add who you are, who your clients are, and how you like answers written. That is now the brief Claude reads every session (Claude Code memory docs).
  4. Install Claude Code and log in with your Claude account. The help center walks through it: install, then "log in with the same credentials you use for Claude" (Claude Help Center). If you prefer clicking to typing, the desktop app is one of the supported surfaces (Claude Code docs).
  5. Open the folder and ask for one real job. Not "hello". Something like: "Read the notes in clients/acme and draft this week's update email into clients/acme/drafts." Watch what it does. Correct it. Put the correction in CLAUDE.md.
  6. Keep your Projects for what they are good at. Quick thinking on your phone, a chat with no files involved. Nothing says you must use one or the other.

Give it a week of real jobs before you judge it. The first day feels slower. By the end of the week the folder knows more about your business than any single Project did.

Where the AI-OS fits, honestly

Full disclosure: I built the AI-OS. It is a one-time setup that turns Claude Code into an operating system for your work. One folder on your own computer, organised in five layers: skills (what it can do), tools (what it can reach), automations (what it runs alone), context (what it knows about you) and knowledge (what it remembers). I run my app, my studio and my content on it.

It is not a replacement for either product in this article. It runs on your own Claude subscription, which you pay Anthropic for separately, and Claude Code runs inside it. What it gives you is the structure from step 1 to 6 above, done properly, so you are not inventing a folder layout and a CLAUDE.md from scratch. It is plain text files on your computer, it asks before it edits, and you can leave anytime and keep everything.

When it does not fit: if Projects already cover your work, you do not need it. If you just want to try Claude Code for a weekend, plain Claude Code and the steps above are enough. The AI-OS starts to earn its place once you have decided Claude Code is where your work lives and want it organised for the long run. If you want the bigger picture of what that means, read what is an AI operating system.

The verdict

Start with the job, not the tool. If the job ends in "and then I paste it into my document", try Claude Code.

Sources