# How to use Claude Projects, with artifacts and memory

> To use Claude Projects, open Projects in the sidebar, click + New Project, add files to its knowledge and set project instructions. Every chat in that project reads both. Artifacts hold what Claude makes, such as documents and small tools, in a panel beside the chat. Memory carries details between chats, and each project keeps a memory of its own.

Mindket editors · Published 2026-11-05 · Updated 2026-11-05 · https://mindket.com/learn/ai-assistants/claude-projects

## In short

- Write a decision into the project's files, or the next chat starts without it.
- Leave the description for yourself and give Claude its brief in the instructions.
- Keep one client to a project, so their files and memory never mix.
- Send a finished document straight to Word or Google Docs instead of copying it out.

## Before you start

You need a Claude account, on the web, the desktop app or the phone. Projects come with every plan, the free one included.

Two terms come up. **Knowledge** is the set of files you add to a project, and Claude uses them in every chat there. The **context window** is how much text Claude can take in at once. On paid plans, when a project's knowledge nears that limit, Claude turns on RAG, short for retrieval-augmented generation, a way of looking up the passages a question needs instead of reading everything. [Anthropic's projects page](https://support.claude.com/en/articles/9517075-what-are-projects) says it expands capacity by up to 10 times.

The steps and menu names follow Anthropic's help center as it read on 5 November 2026. If you want the mechanics of retrieval, [the RAG tutorial](https://mindket.com/learn/rag/rag-tutorial) builds a small version by hand.

## Set up a project in five steps

Pick one job you repeat, such as a newsletter or a client's weekly report. The instructions below are our own example for a small bakery. Swap in your business and keep the shape: who reads it, the tone, and where the facts come from.

````text
You help me write the monthly email newsletter for a small bakery.
Readers are regular customers. Keep a warm, plain tone and paragraphs of two or three sentences.
Take prices and opening hours only from the files in this project.
If a fact is not in the files, ask me instead of guessing.
````

1. **Open Projects.** Find it on the left side of Claude, or go to [claude.ai/projects](https://claude.ai/projects), and click + New Project in the upper right.
2. **Name it and describe it.** Both are for you. Claude cannot see the name or the description, so nothing it needs goes there.
3. **Write the instructions.** Click Set project instructions, paste your text and click Save instructions.
4. **Add knowledge.** In the knowledge base on the right of the project page, click + and upload documents, text files or code.
5. **Chat inside the project.** Start a new chat from the project page. To bring in an older chat, click the dropdown arrow next to its name, choose Add to project and pick the project in the Move chat window.

## What Claude reads in a project

Chats inside one project do not share what was said in them. Agree on a new price in Monday's chat, and Tuesday's chat will not know it unless you add it to the knowledge base. Keeping those files current is most of the upkeep a project needs.

**Memory** fills part of that gap. Claude saves it as separate topics while you chat, and each project has its own. A project also draws on the memory from your chats outside any project, never on another project's. For a client whose details must stay sealed, turn off Use account memory on that project's page.

Memory is on by default on the Free, Pro, Max and Ultra plans. On Team and Enterprise, where the owners decide whether it is available, it stays off for each member until they turn it on. [Anthropic's memory page](https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context) puts the list under Settings, Your Claude: select Manage next to Your memory, then edit or delete any topic.

**Projects and memory in Claude, by date**

1. **June 25, 2024** Projects. Chats can be grouped into shareable projects, each with its own large context window.
2. **August 1, 2025** Project sharing. Permissions and sharing arrive for Team and Enterprise plans.
3. **October 23, 2025** Memory on Max and Pro. Max users get memory first, and Pro follows over about two weeks.
4. **March 2, 2026** Memory for everyone. Memory from chat history reaches every Claude user, free accounts included.
5. **August 25, 2026** Memory in Cowork. Memory works across chat and Cowork in the cloud, and you can edit or delete each topic.
6. **September 16, 2026** Cowork in every chat. What Cowork does becomes available from any conversation, with your projects carried over. It rolls out gradually to Pro and Max.

Sources: [Anthropic (@AnthropicAI) on X: Projects launch](https://x.com/AnthropicAI/status/1805616725733339199) (read 2026-11-05), [Anthropic: Claude release notes](https://support.claude.com/en/articles/12138966-release-notes) (read 2026-11-05)

| Layer | Where to set it | Reaches | Use it for |
| --- | --- | --- | --- |
| Instructions for Claude | Your initials at the lower left, then Settings, Your Claude | Every conversation | How you like answers, the terms you use |
| Project instructions | Set project instructions, on the project page | Chats in that project | The job's readers, tone and rules |
| Project knowledge | The + button in the knowledge base | Chats in that project | Price lists, past reports, style samples |
| Memory | Saved by Claude as you chat, managed in Settings, Your Claude | Your chats, with one memory per project | Details you would otherwise repeat |

## Artifacts: what Claude makes for you

When Claude writes something long and self-contained, typically over 15 lines, it opens beside the chat as an **artifact**. Anthropic defines one as anything Claude makes for you that you would put in front of someone: a document, a diagram, a dashboard, a one-page website or a small interactive tool.

They need one setting, Cloud code execution and file creation, under Settings, then Capabilities. Every plan has them, Free included, and each one starts private to you.

To change one, ask Claude, or highlight text in a document, click Edit with Claude and describe the change. Export sends a document to Word, PDF, Markdown or Google Docs, and a slide deck to PowerPoint, PDF or Google Slides.

Some artifacts call Claude themselves and work as small apps. Each person who opens one signs in with their own Claude account and spends their own plan's usage, so you never hand out an API key.

## How Anthropic introduced Projects on X in 2024

Anthropic [introduced Projects on X](https://x.com/AnthropicAI/status/1805616725733339199) on 25 June 2024. The post made two promises: chats grouped into projects you could share, and a 200K context window in each, with room for the documents, code and files the work needs.

Both have grown since. [Claude's release notes](https://support.claude.com/en/articles/12138966-release-notes) record project permissions and sharing for the business plans from August 2025. The window is no longer the ceiling on paid plans either, because retrieval takes over when a project's knowledge nears the limit.

What has not changed is the idea of one project per job. Give each client or report a project of its own, and keep its knowledge to the files that job needs.

> You can now organize chats with Claude into shareable Projects.
>
> Anthropic, the AI safety and research company that makes Claude, on its own account, [X](https://x.com/AnthropicAI/status/1805616725733339199) (read 2026-11-05)

[![Anthropic (@AnthropicAI) on X, 25 June 2024: You can now organize chats with Claude into shareable Projects. Each project includes a 200K context window, so you can include relevant documents, code, and files.](https://mindket.com/blog/shots/x-anthropicai-projects-launch.webp?v=1b881c18941e)](https://x.com/AnthropicAI/status/1805616725733339199)

Anthropic on X, 25 June 2024

> A screenshot of X's own embed of the post, taken on 5 November 2026, with the time in UTC. Select it to open the post on X.

## Common mistakes

- Leaving last year's price list in the knowledge base. Claude uses every file there, so remove the old version when you upload a new one.
- Switching memory off once a chat is under way. The Memory switch in the + menu works only before your first message.
- Asking something private in a normal chat. An incognito chat, from the ghost icon at the upper right, stays out of your history and your memory.

## Try it yourself

- Build the bakery project with your own business in it, then ask for next month's newsletter.
- In a second chat, ask about something you settled in the first one. If Claude has no idea, put that fact in a knowledge file and ask again.
- Read the topics Claude has saved about you, and delete one that is out of date.
- Export one artifact to Word and look at what came through.

## Next: NotebookLM

Claude answers from your project files and from its own training. [The NotebookLM tutorial](https://mindket.com/learn/ai-assistants/notebooklm-tutorial) works the other way round: Google's notebook answers only from the sources you add and quotes the passage behind each claim, which makes checking a fact a single click.

## Check yourself

1. Where do you put the rules every chat in a Claude project should follow?

   A. In the project description
   B. In Set project instructions
   C. In the project name
   D. In a chat you pin to the top

   Answer: B. Claude cannot see a project's name or description, while project instructions apply to every chat in that project.

2. You agreed a new price in one chat of a project, and the next chat there does not know it. Why?

   A. Chats in a project do not share what was said unless it is in the knowledge base
   B. Memory works only on paid plans
   C. The project was archived

   Answer: A. Context is not shared across project chats, so a decision has to be added to the knowledge base to reach the next chat.

3. A client's project should ignore what Claude remembers from your other chats. What do you turn off?

   A. Search and reference chats
   B. Cloud code execution and file creation
   C. Edit with Claude
   D. Use account memory, on that project's page

   Answer: D. A project draws on your non-project memory unless Use account memory is off on its page.

4. You want the newsletter Claude wrote as a Word file. What do you click?

   A. Share project
   B. Set project instructions
   C. Export

   Answer: C. Export sends a document artifact to Word, PDF, Markdown or Google Docs.

## Frequently asked questions

### Are Claude Projects free?

Yes, within a limit. A free account can create up to five projects. Paid plans also get the retrieval mode that keeps a large knowledge base usable.

### Can I share a Claude project with my team?

Sharing comes with Team and Enterprise. Open the project, click Share project, add people by name or email and give each Can view or Can edit, then click Share. Can edit lets them change the instructions and knowledge and manage members.

### What goes in project instructions and what goes in knowledge?

Rules go in the instructions, material in the knowledge. "Short paragraphs, ask before guessing" is an instruction, and the price list is knowledge.

### Can I delete a Claude project I archived?

Not while it is archived. Unarchive it first from the three dots menu, then choose Delete and Yes, delete. Archiving keeps the chats, the members and the sharing settings as they were.

## Sources

- [Anthropic: What are projects?](https://support.claude.com/en/articles/9517075-what-are-projects) (read 2026-11-05)
- [Anthropic: How can I create and manage projects?](https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects) (read 2026-11-05)
- [Anthropic: What are artifacts and how do I use them?](https://support.claude.com/en/articles/9487310-what-are-artifacts-and-how-do-i-use-them) (read 2026-11-05)
- [Anthropic: Use Claude's chat search and memory to build on previous context](https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context) (read 2026-11-05)
- [Anthropic: Understanding Claude's personalization features](https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features) (read 2026-11-05)
- [Anthropic (@AnthropicAI) on X: Projects launch](https://x.com/AnthropicAI/status/1805616725733339199) (read 2026-11-05)
- [Anthropic: Claude release notes](https://support.claude.com/en/articles/12138966-release-notes) (read 2026-11-05)

[Browse MCP servers](https://mindket.com/mcp-servers)