3 Days Free Demo Class — sit in on real classes before you enrol.Book your free demo → 9991 729 730

Claude Projects Explained: What They Are and How to Use Them

Claude Projects are the easiest way to stop explaining yourself to Claude again and again. Instead of pasting the same brief, style guide and background files into every new chat, you put them in one Project once. Every chat you start inside that Project already knows them. This guide covers what a Project is, how to set one up, how it affects your usage, who can share one, and when a skill or a simple saved prompt is the better tool.

Riya Malik 7 min read Updated November 2026
Claude Projects Explained: What They Are and How to Use Them

What are Claude Projects?

A Claude Project is a self-contained workspace inside Claude. Anthropic describes each one as having its own chat history and its own knowledge base. Think of it as a labelled folder on your desk: the brief, the notes and the rules for one job all live inside it.

Projects are one of several ways to shape how Claude works. Skills are another, and our guide to what Claude skills are explains that side. Every Project has three main parts:

  • Project knowledge:: files and text you upload, such as PDFs, Word documents, spreadsheets, notes or code. Claude uses them as background in every chat in the Project.
  • Project instructions:: a short note that tells Claude how to behave in this Project, for example the tone to use or the role to play.
  • Chats:: the conversations you start inside the Project. Each one sees the knowledge and instructions automatically.
Diagram of a Claude Project called Client A containing three parts: project knowledge, project instructions and chats, with its own separate memory
A Project bundles what Claude should know, how it should behave, and the chats that use both.

How do Claude Projects work?

When you open a new chat inside a Project, Claude starts with the Project's instructions and knowledge already in view. You just ask.

Chats inside the same Project do not read each other. If you settle something important in one chat, add it to the project knowledge so the next chat knows it too.

The Project's name and description are for you. Anthropic notes that Claude cannot see them, so anything Claude must know belongs in the instructions or the knowledge.

Memory is kept per Project too. Inside a Project, Claude uses that Project's memory plus your general account memory, but never memory from your other Projects. If you want a Project fully sealed off, turn off the 'Use account memory' option on the Project's page.

How to use Claude Projects: set one up step by step

You can do this in the Claude web app in about five minutes. These steps follow Anthropic's own help centre.

  1. Click Projects in the left sidebar, or go straight to claude.ai/projects.
  2. Click '+ New Project' in the top right corner.
  3. Give it a clear name and a one-line description so you can find it later. On Team and Enterprise plans you also choose who can see it.
  4. Click 'Set project instructions', write your guidance, then click 'Save instructions'.
  5. In the project knowledge area on the right, click the '+' button and upload your files or add text.
  6. Start a chat inside the Project and ask a test question, such as 'Summarise the brief in five bullet points.' If the answer uses your files, the setup works.
Five numbered steps to set up a Claude Project: open Projects, click New Project, name it, set project instructions, add files to project knowledge
Five steps, then a quick test chat to check Claude is using your files.

What should you put in project knowledge and instructions?

Keep the two jobs separate. Knowledge is what Claude should know. Instructions are how Claude should behave. Good uses for a Project include one per client, one per course you teach, or one for a long report.

Below is a full set of project instructions for a made-up cafe client. Copy it and change the details. For help writing your own, the same rules that make a good chat prompt apply, and our simple prompt formula for Claude works well here.

You are the content assistant for GreenLeaf Cafe, a vegetarian cafe.
Audience: college students and young professionals.
Tone: warm, simple, no jargon.
Follow brand-guide.pdf for words to use and avoid.
Keep Instagram captions under 60 words.
If a detail is not in the project files, ask me instead of guessing.
  • Good knowledge:: the client brief, brand or style guide, product details, past examples you liked, FAQs and research notes. In short, material Claude will refer to again and again.
  • Good instructions:: who Claude is in this Project, who the audience is, the tone, format rules, and what to do when information is missing.
  • Leave out:: old drafts you no longer trust, duplicate copies of the same file, and anything unrelated to this Project. Extra files make answers less focused and add to what Claude reads.

Do Claude Projects use more tokens?

Short answer: a Project with lots of files can use more of your limit per message, but it is usually cheaper than uploading the same files again in every chat.

Here is why. Your project knowledge is loaded as background for every chat in the Project, so a bigger knowledge base means more for Claude to read with each message. The good news, from Anthropic's usage guide, is that Project content is cached, and cached content counts less against your limits when it is reused. The catch is that caches expire after a period of inactivity, so the first message after a long break counts in full again.

Very large Projects work differently on paid plans (Pro, Max, Team and Enterprise). When project knowledge gets close to what Claude can hold at once, Claude switches on retrieval, called RAG, automatically. Instead of reading every file, it searches your knowledge and pulls in only the relevant parts. Anthropic says this lets a Project hold up to 10 times more content.

Tip: upload your core documents once, remove files you no longer need, and keep related questions in the same Project.

Flow chart: in a normal Claude Project files are read in every chat and cached; in a very large Project on paid plans RAG switches on and Claude searches only the relevant parts
Caching makes reused Project files cheaper; very large Projects switch to search instead.

Claude Projects limits and plan availability

Anthropic is testing a new version of Projects. In this beta, a Project is one ongoing conversation where Claude splits work into parallel threads that run in the cloud. It is reaching select Pro and Max users in Claude Code first, and chat Projects stay on the current version for now.

Here are the current limits for chat Projects, from Anthropic's help centre:

  • Who can use them:: all users, including the free plan. Free users can create up to five Projects.
  • File size:: up to 30MB per file in project knowledge.
  • Number of files:: no fixed cap, but the total content must fit within Claude's context window. On paid plans, RAG takes over when it gets close.
  • File types:: PDF, DOCX, CSV, TXT, HTML, ODT, RTF, EPUB, JSON and XLSX, plus JPEG, PNG, GIF and WebP images. XLSX needs code execution and file creation turned on in settings.
  • Sharing:: Team and Enterprise plans only, covered below.

Can you share Claude Projects or make one public?

Project sharing is available on Team and Enterprise plans only. There you can keep a Project private or make it public, but 'public' means everyone in your organisation, not the whole internet. To share, click 'Share project' next to the Project name, add people, and pick 'Can view' (see the Project and chat in it) or 'Can edit' (change instructions and knowledge, and manage members).

Your own chats stay private even inside a shared Project, unless you share a chat yourself.

On Free, Pro and Max there is no way to share a Project, and Anthropic does not describe any way to publish a Project on the open web. What you can do is share a single chat as a snapshot link. Anyone with the link sees the messages up to the moment you shared it. On Team and Enterprise plans, chat links only work inside your organisation.

Comparison card: on Free, Pro and Max you cannot share a Claude Project but can share one chat link; on Team and Enterprise you can share with Can view or Can edit access within your organisation
Project sharing is a Team and Enterprise feature, and 'public' stays inside your organisation.

Is it possible to recover deleted projects on Claude?

Anthropic's help centre explains how to delete a Project: open the three dots menu, choose 'Delete', then confirm with 'Yes, delete'. It does not describe any way to restore a Project after that. Treat deleting as permanent.

If you might need a Project again, archive it instead. Archiving keeps the knowledge, members and sharing settings. Archived Projects sit in the archived projects tab, and you can bring one back through the three dots menu with 'Unarchive'. An archived Project cannot be deleted until you unarchive it, which adds a useful safety step.

Claude Projects vs skills vs a saved prompt: when to use which

These three tools solve different problems. Anthropic sums it up well: a Project gives Claude what it needs to know, while a skill teaches Claude how to do something.

  • A saved prompt:: best for a one-off or occasional request. You paste it into a chat when needed. Nothing loads by itself.
  • A Project:: best for one ongoing piece of work with its own files, such as a client, a course or a report. Its knowledge is always present in that Project's chats, and nowhere else.
  • A skill:: best for a task you repeat the same way everywhere, such as writing in your brand voice or following a review checklist. Claude loads it only when the task calls for it, in any chat or Project, and it can include scripts and templates.
Three cards comparing a saved prompt, a Claude Project and a Claude skill by what each is best for and when each one loads
Prompts for one-offs, Projects for one ongoing job, skills for tasks you repeat.

A simple rule for choosing

If you keep retyping the same prompt, or copying the same instructions into several Projects, turn it into a skill. Our guide on how to create your own Claude skill walks you through it without any code.

The two also work well together. The Project holds a client's files, and the skill holds your method. If you want guided practice with Claude and other AI tools for marketing work, see our AI tools course.

Riya Malik
Written by Riya MalikContent Writer, Digihandler Academy

Riya writes Digihandler Academy's guides and roadmaps. Every article is fact-checked and reviewed by Lovish Madaan, founder & trainer, who has run the Digihandler agency since 2018. More from Riya →

The Claude skills series

  1. Part 1: What are Claude skills?
  2. Part 2: How to install Claude skills
  3. Part 3: How to create a Claude skill
  4. Part 4: How to prompt Claude and Claude Code
  5. Part 5: Claude Projects explained (you are here)

Common questions

A Claude Project is a workspace inside Claude for one ongoing job, such as a client, a course or a report. You upload files to its project knowledge and write project instructions once. Every chat you start inside the Project can then use those files and follow those instructions automatically, so you do not need to paste the same background into each new conversation. Each Project also keeps its own chat history and its own memory, separate from your other Projects.

Yes. According to Anthropic's help centre, Projects are available to all users, including the free plan, but free users can create a maximum of five Projects. Paid plans (Pro, Max, Team and Enterprise) add an automatic retrieval mode for very large Projects, which lets a Project hold up to 10 times more content. Sharing a Project with colleagues is limited to the Team and Enterprise plans.

Project knowledge is loaded as background for every chat in the Project, so a large knowledge base means more content per message. However, Anthropic says Project content is cached and counts less against your usage limits when it is reused, which usually makes it cheaper than uploading the same files again in each chat. Caches expire after inactivity. On paid plans, very large Projects switch to retrieval, where Claude searches for only the relevant parts.

Not to the open internet. Anthropic's help centre only describes Project sharing on Team and Enterprise plans, and there 'public' means visible to everyone in your organisation. You can also invite specific people with 'Can view' or 'Can edit' access. On Free, Pro and Max you cannot share a Project at all, but you can share a single chat as a snapshot link that anyone with the link can open.

Anthropic's help centre explains how to delete a Project but does not describe any way to restore one, so treat deletion as permanent. If you think you may need a Project later, archive it instead. Archiving keeps its knowledge, members and sharing settings, and you can unarchive it any time from the archived projects tab. Before deleting, download any files you want to keep and save a copy of your project instructions.

More from the blog

AI in digital marketing: what actually works right now

AI in digital marketing: what actually works right now

Where AI genuinely saves marketers time on real client work, and where it quietly makes things worse.

Read article →
AI SEO: how to get cited inside AI search results

AI SEO: how to get cited inside AI search results

What decides whether an AI assistant quotes your page, and the fixes worth doing first.

Read article →

Learn this properly

3 days free · Rohtak campus & live online

We'll call you within 24 hours. No spam, ever.

Prefer to talk? Call 9991 729 730.