Claude Projects: What They Are, How to Set One Up, and 5 Setups for Coaches

Most operators use Claude like a stranger. Every chat starts from zero, and every chat starts with the same five paragraphs of context.

Claude Projects fix that. A project holds your documents and your standing instructions, so every chat inside it starts already briefed. Set one up properly and the repeated work of a coaching business, from content to onboarding to offers, stops depending on how well you explain yourself that morning.

Below: what Claude Projects are as of October 2026, which plans include them, how to set one up step by step, and five project setups built for coaches and founders. Every feature described here is checked against Anthropic's own help center. Features change, so treat the help center as the final word.

What Claude Projects are

Anthropic's definition is short: "Projects allow you to create self-contained workspaces with their own chat histories and knowledge bases" (Claude Help Center: What are projects?).

Each project has two working parts.

• Project knowledge. Documents, text, code or other files you upload. Claude uses them as context in every chat inside that project.
• Project instructions. Standing guidance on how Claude should behave and respond in that project. Anthropic's examples: a more formal tone, or answering from the perspective of a specific role or industry.

Everything else is organisation: chat history scoped to the project, starring, archiving, and on team plans, sharing.

Which plans include Claude Projects

Several older guides still say Projects require a paid plan. That is out of date. Per the help center, "Projects are available to all users, including those with free Claude accounts. Free users can create a maximum of five projects."

What paid plans add:

• Expanded knowledge with RAG (Pro, Max, Team and Enterprise). When your project knowledge approaches the context window limit, Claude automatically switches to retrieval augmented generation and searches your files instead of loading all of them at once. Anthropic says this expands capacity "by up to 10x" (RAG for projects). No setup is required.
• Sharing (Team and Enterprise only). Projects can be shared with specific members or the whole organisation, with two permission levels: "Can view" and "Can edit."

Projects and memory

Two details are easy to miss and matter in practice.

First, chats inside a project do not automatically see each other. The help center is explicit: "Context is not shared across chats within a project unless the information is added into the project knowledge base" (How can I create and manage projects?). If a decision matters for future chats, write it into the knowledge base.

Second, where Claude's memory is on, each project has its own memory, kept separate from your non-project chats. Memory is on by default on Free, Pro and Max plans; on Team and Enterprise it depends on whether an owner has enabled it. You can turn memory off for a single chat from the "+" menu before sending the first message.

The new version of Projects (beta)

Anthropic is rolling out a new version of Projects in stages, starting with Claude Code. In it, a project is one conversation that Claude breaks into parallel threads running in the cloud, and a Library collects your files and the files Claude produces. As of today it is in beta for select Pro and Max subscribers who use Claude Code, with chat, Cowork, Team and Enterprise to follow (help center).

Existing projects in chat keep working as they do now. Everything in the setup guide below describes that current version.

How to set up a Claude project, step by step

The mechanics take two minutes. The thinking takes longer and is where the value is.

1. Open Projects. Go to claude.ai/projects, or open the left side of your account and click "Projects."
2. Create the project. Click "+ New Project" in the upper right. Give it a name and description. Note: Claude does not have access to the name and description, so do not put instructions there.
3. Set visibility (Team and Enterprise only). Keep it private, or share it with your organisation where admins allow it.
4. Write the project instructions. Click "Set project instructions," write how Claude should behave in this project, and click "Save instructions." These apply to every chat in the project.
5. Add project knowledge. In the knowledge base on the right side of the project page, click "+" and upload the documents, text files or code that Claude should treat as context.
6. Name files clearly. Anthropic's RAG guidance recommends clear, descriptive filenames and referencing specific documents by name when you ask questions. Do this from day one; it costs nothing.
7. Start chats inside the project. Existing standalone chats can be moved in with the dropdown next to the chat name and "Add to project."
8. Maintain it. Star the projects you use daily. Archive the ones you are finished with; archived projects keep their knowledge and members. To delete an archived project, unarchive it first.

What good project instructions look like

Instructions are the standing brief. Write them the way you would brief a capable new hire on day one.

• Role: who Claude is in this project and who the output is for.
• Standards: voice, format, length, and what "done" looks like.
• Sources: which knowledge files are authoritative for what.
• Boundaries: what Claude should not do, such as inventing client results, quoting prices that are not in the files, or making claims you cannot back.

Keep instructions about behaviour. Keep facts in knowledge files, where you can update them without rewriting the brief.

Five Claude project setups for coaches and founders

One project per repeated job. Not one giant project for the whole business. Narrow projects keep the knowledge relevant and the instructions sharp.

1. The content engine

Job: turn your ideas into posts, emails and articles in your voice.

• Knowledge: a voice guide, a list of words you never use, your offer summary, and a handful of your best past pieces as examples.
• Instructions: write in the voice guide, match the format of the examples, flag any factual claim that needs a source, never invent client stories.
• Use it for: drafting, repurposing one long piece into several short ones, and building monthly plans. The workflow in writing a month of LinkedIn content with Claude runs well inside a project like this.

2. Client onboarding

Job: produce consistent welcome material, intake summaries and first-week plans.

• Knowledge: your onboarding checklist, intake questionnaire, program outline, policies, and template welcome emails.
• Instructions: follow the checklist in order, use only the policies in the files, and mark anything missing from a client's intake rather than filling the gap.
• Use it for: drafting each new client's welcome pack and summarising their intake into a first-week plan for you to review.

Treat client data with care. Upload only what the task needs, and check your plan's data settings against your own privacy obligations before adding client information.

3. Offer development

Job: pressure-test and refine what you sell.

• Knowledge: your current offer, pricing, a summary of your ideal client, notes from sales calls, and objections you hear repeatedly.
• Instructions: act as a skeptical buyer first, then as a strategist; separate what the files support from what is assumption.
• Use it for: stress-testing a new program, drafting sales page sections, and rehearsing objections. For a worked example of building an offer this way, see using AI for offer creation.

4. The weekly review

Job: turn a week of raw notes into decisions.

• Knowledge: your quarterly goals, your review template, and a running decisions log.
• Instructions: compare the week against the goals, surface what slipped and why, and propose no more than three priorities for next week.
• Use it for: one chat per week. Because chats do not share context, paste each week's outcome back into the decisions log file so the next review can see it. Pair it with the Stoic weekly review for the reflective half.

5. Strategy and decisions

Job: think through bigger calls with your full context loaded.

• Knowledge: the business plan, key numbers you are willing to share, past decisions and their outcomes.
• Instructions: argue against the user's preferred option before supporting it, list the assumptions each recommendation depends on, and say plainly when the files do not contain enough information.
• Use it for: pre-mortems, pricing decisions and hiring calls. The prompts in 10 Claude prompts for business strategy work as starting points inside this project.

Common mistakes with Claude Projects

• Putting instructions in the description. Claude cannot see the project name or description. Instructions go in project instructions.
• Expecting chats to remember each other. They do not, unless the information is in the knowledge base. Write important outcomes back into a file.
• One project for everything. A single project stuffed with every document dilutes relevance. Split by job.
• Stale knowledge. An outdated price list or old offer in the files produces confidently wrong output. Review knowledge files when the business changes.
• Vague filenames. "doc1.pdf" helps nobody. Anthropic recommends descriptive names, and it makes your own maintenance faster too.
• Shipping output unchecked. A well-briefed model can still be wrong. Verify facts, numbers and claims before anything reaches a client. The checks in verifying delegated AI work apply here too.

Frequently asked questions

Are Claude Projects free?

Yes. According to Anthropic's help center, Projects are available to all users, including free accounts, with free users limited to a maximum of five projects. Paid plans (Pro, Max, Team and Enterprise) add expanded project knowledge through RAG, and Team and Enterprise add project sharing with view and edit permissions.

What is project knowledge in Claude?

Project knowledge is the set of documents, text, code and other files you upload to a project. Claude uses them as context in every chat inside that project. On paid plans, when knowledge approaches the context window limit, Claude automatically switches to RAG and searches the files instead of loading them all.

Do chats in a Claude project share context?

Not automatically. Anthropic states that context is not shared across chats within a project unless the information is added to the project knowledge base. Where memory is enabled, each project also has its own memory, separate from non-project chats. Save important decisions as a knowledge file to carry them forward.

Can I share a Claude project with my team?

Only on Team and Enterprise plans. Project creators can share with individual members by email, add several people at once, or make the project visible to the whole organisation where admins allow it. Members get either "Can view" access, which allows chatting, or "Can edit" access to change instructions and knowledge.

How many Claude projects should a coach run?

One per repeated job. A useful starting set is a content engine, client onboarding, offer development, a weekly review and a strategy project. Free accounts are capped at five projects, which happens to fit that set. Keep each one narrow so its knowledge stays relevant and its instructions stay specific.

Systems compound. Prompts do not.

A good project is a system you build once and use every day. The same principle runs the body and the mind. The free 5-Day Stoic Operator Challenge applies it to training, Stoic practice and the daily rhythm that makes tools like this worth having.