What you will be able to do
- Explain what a Claude Project is and what it gives each chat inside it
- Create a project and know which details Claude can and cannot see
- Add knowledge sources so that every chat in a project has the same background
- Describe how RAG lets a project hold more knowledge, and what you need to do about it (nothing)
Key concept
Project-scoped context — A project is a container whose knowledge base, and whose instructions, apply to every chat started inside it. Anything outside that container, including what was said in another chat, is not carried over automatically.
1.What a project is
Suppose you use Claude every week for the same piece of work: a client engagement, a newsletter, a quarterly review. In ordinary chats you would paste the same brand guide, the same background and the same 'please write formally' into every new conversation. A Claude Project exists to stop that repetition. The documentation describes projects as self-contained workspaces with their own chat histories and knowledge bases. Inside a project you upload documents, provide context, and hold focused chats with Claude.
Two parts of a project carry over into every chat inside it. The knowledge base holds the material Claude should know about: documents, text, code and other files. The project instructions say how Claude should behave: tone, role, perspective. This page covers the knowledge side. Instructions and sharing have their own lesson.
Projects are not only for paying customers. They are available to all users, including free accounts, but free users can create at most five projects. That cap is the plan boundary for projects themselves that you should remember.
Checkpoint 1 of 6· Check yourself
A user on a free Claude account asks whether they can use Projects at all. What is accurate?
Projects are available to every user. The only restriction stated for free accounts is a maximum of five projects.
“Free users can create a maximum of five projects.”Source: support.claude.com
2.Creating a project, and what Claude can see
To create a project, open the Projects section (hover over the left side of your account and click “Projects,” or go to claude.ai/projects). Click "+ New Project" in the upper right corner and give the project a name and description. On a Team or Enterprise plan you also choose visibility: keep it private to you and the members you invite, or share it with your broader organization if administrators allow that. Once the project is set up, you can start chats inside it.
One detail in this flow is a common source of mistakes. The name and description are labels for you. Claude cannot see them. A description such as "Formal tone, Q3 board audience" changes nothing about how Claude responds. Anything Claude should act on has to go in the project instructions or the knowledge base.
| Project element | Does Claude use it? | Scope |
|---|---|---|
| Name and description | No. Claude will not have access to these details | Labels for people only |
| Project knowledge base | Yes, used as context | All chats within the project |
| Project instructions | Yes, used to shape behaviour | All chats within the project |
| What was said in one chat | Not in other chats | Only that chat, unless added to the project knowledge base |
Checkpoint 2 of 6· Put it in order
Put the steps for starting work in a new project in order
- 1.Click "+ New Project"
- 2.Initiate chats with Claude within the project
- 3.Give the project a name and description
- 4.Open the Projects section (claude.ai/projects)
You create the project and label it first. Chats come after the project exists.
“Once your project is set up, you can initiate chats with Claude within the project.”Source: support.claude.com
Sources2
3.Adding knowledge sources
The project knowledge base is on the right side of the project's main page. Click the "+" button and upload relevant documents, text files or code snippets. Claude processes this material and uses it as context in every chat in the project. You can add documents, text, code and other files. The point is to give Claude the background for each chat, so you don't have to re-supply it.
Compare this with attaching a file to a single chat. A file attached in a chat is available in that conversation only. A file uploaded to the project is there for every conversation in the project. The general rule from the upload documentation is that files can go to individual chats, or to a project for persistent reference across conversations.
This is also why the predict above has the answer it does. Chats in a project share the knowledge base, but they do not share each other's conversations. If a fact comes up in one chat and should apply from now on (a decision, a corrected figure, a new rule), put it in the project knowledge. Do not assume Claude will remember it in the next chat.
Checkpoint 3 of 6· Check yourself
A manager wants every newsletter draft in a project to draw on the same brand guide and past issues. Where should those files go?
Material in the knowledge base is used across all chats in the project. A chat attachment stays in that chat, and Claude cannot see the description.
“Anything you upload to this space will be used across all of your chats within that project.”Source: support.claude.com
Checkpoint 4 of 6· Exam question
What is the primary purpose of the knowledge base within a Claude Project?
Correct answer: A — It gives Claude uploaded documents, text, and code as shared context for every chat inside the project
- A. Correct. The project knowledge base holds documents, text, code, and other files that Claude uses as background context across all chats within that project.
- B. Incorrect. The knowledge base is not a prompt history log; it is a repository of uploaded reference material scoped to one project.
- C. Incorrect. Files uploaded to a project's knowledge base are not automatically published outside the project; access depends on project sharing settings.
- D. Incorrect. The knowledge base supplies context to Claude within the project; it does not spin up a separate chatbot.
4.When knowledge grows: RAG mode
Normally, project knowledge is loaded into Claude's context window. That used to set a hard ceiling: once a project reached the context window's capacity, you could not add more. Retrieval augmented generation (RAG) removes that ceiling.
When project knowledge approaches or exceeds the context window limit, Claude automatically switches the project to RAG mode, which expands capacity by up to 10x. In RAG mode Claude does not load everything at once. It uses a project knowledge search tool to retrieve the most relevant parts of your uploaded documents, and you will see it using that tool. The project also shows a visual indicator that it is RAG-enabled. If the knowledge later drops back below the threshold, Claude can switch back to in-context processing. The documentation says that, when possible, projects use in-context processing for optimal performance.
You don't configure any of this. Activation is automatic and needs no setup. RAG also works with web search, extended thinking and Research. The documentation says response quality stays consistent with in-context processing.
What you can do is make retrieval easier. Name files well, keep related documents in the same project, and refer to a specific document by name when you want Claude to focus its search there.
Checkpoint 5 of 6· Check yourself
A project's knowledge base has grown close to the context window limit. What does the project owner need to do to keep adding documents?
RAG switches on by itself when knowledge nears the limit, with no setup or configuration.
“Seamless transition: Automatic activation when needed, no setup required.”Source: support.claude.com
Checkpoint 6 of 6· Check yourself
Which practice does the documentation recommend for getting better answers from a large, RAG-enabled project?
Good file names help retrieval. The documentation also recommends grouping related documents in one project and naming specific documents in questions.
“Well-named files help Claude understand and retrieve the right information more effectively.”Source: support.claude.com
Sources4
Exam traps
Each one states something that sounds right. Open it to see what is actually true.
1.Putting guidance in the project description ("formal tone, board audience") will shape Claude's answers.Why is that wrong?
Claude cannot see the project name or description. Guidance has to go in the project instructions, and reference material in the knowledge base.
Covered in Creating a project, and what Claude can see
2.Because chats sit in the same project, a fact established in one chat is known in the next.Why is that wrong?
Chats in a project share the knowledge base and instructions, not each other's conversations. Anything that should persist has to be added to project knowledge.
Covered in Adding knowledge sources
3.A large knowledge base needs RAG to be switched on or configured.Why is that wrong?
RAG turns on automatically as knowledge approaches the context limit. It needs no setup and can switch back when the knowledge shrinks.
Covered in When knowledge grows: RAG mode
Sources
Every claim above is drawn from one of these pages, quoted as it was written on the date shown.
- 1.
“Projects allow you to create self-contained workspaces with their own chat histories and knowledge bases.”
↩︎ What a project is“Existing projects in chat and Cowork keep working as they do today.”
↩︎ What a project is“You can upload relevant documents, text, code, or other files to a project's knowledge base”
↩︎ Adding knowledge sources - 2.
“Free users can create a maximum of five projects.”
↩︎ What a project is“Give your project a name and description (note that Claude will not have access to these details).”
↩︎ Creating a project, and what Claude can see“Share with your broader organization: Everyone in your organization can view and use this project (where enabled by administrators).”
↩︎ Creating a project, and what Claude can see“Upload relevant documents, text files, or code snippets.”
↩︎ Adding knowledge sources“Anything you upload to this space will be used across all of your chats within that project.”
↩︎ Key concept“Give your project a name and description (note that Claude will not have access to these details).”
↩︎ Exam trap 1“Context is not shared across chats within a project unless the information is added into the project knowledge base.”
↩︎ Exam trap 2“Once your project is set up, you can initiate chats with Claude within the project.”
↩︎ Checkpoint“Context is not shared across chats within a project unless the information is added into the project knowledge base.”
↩︎ Prediction“Anything you upload to this space will be used across all of your chats within that project.”
↩︎ Checkpoint - 3.
“Files can be uploaded to individual chats or uploaded to a project's Files section for persistent reference across conversations.”
↩︎ Adding knowledge sources - 4.https://support.claude.com/en/articles/11473015-retrieval-augmented-generation-rag-for-projectsOfficial docs
“Claude will automatically enable RAG mode to expand your project's capacity by up to 10x”
↩︎ When knowledge grows: RAG mode“Claude uses a project knowledge search tool to retrieve relevant information from your uploaded documents.”
↩︎ When knowledge grows: RAG mode“If your project knowledge later drops below the context window threshold, Claude can automatically convert back to context-based processing.”
↩︎ When knowledge grows: RAG mode“Yes. RAG works with all Claude features, including web search, extended thinking, and Research.”
↩︎ When knowledge grows: RAG mode“No, RAG activates automatically when needed. No setup or configuration is required.”
↩︎ Exam trap 3“Seamless transition: Automatic activation when needed, no setup required.”
↩︎ Checkpoint“Well-named files help Claude understand and retrieve the right information more effectively.”
↩︎ Checkpoint