Home · Read online · Level 3 — Daily work

Chapter L3.2 — Projects

Level 3 — Daily work. Product details verified on 22/06/2026 against official sources.

Goal

By the end you will know what a Project is, when it’s worth creating one, how to give it instructions and reference files that persist from one conversation to the next, and what changes when you work in a team. We’ll use as an example the Project this book was written with.

Prerequisites

What a Project is

A Project is a workspace dedicated to an activity: it gathers in one place the instructions that always apply, the reference files (the “knowledge base”) and the linked conversations. Every chat opened inside the Project starts already with that context, without you repeating it.

You’d already come across them in passing in the settings (ch. L1.3): here we put them to work. Projects are available on all plans; on the Free plan they are limited to 5. Team sharing is reserved for Team and Enterprise (see the ledger).

When to create one

Not everything deserves a Project. The single question fits fine in a normal chat. It’s worth creating a Project when these signals recur:

If you find yourself pasting the same instructions or the same documents at the start of every chat, it’s time for a Project.

Instructions and knowledge base

Two things make a Project useful:

The difference from CLAUDE.md (ch. L2.4) is the context of use: CLAUDE.md governs Claude Code on the project’s files; the Project instructions govern the conversations inside that Project. Same idea — write the stable context once — applied to different tools.

Note: the knowledge base files have a per-file limit (around 30 MB) but no binding limit on number, as long as the content fits in Claude’s context. It’s a different limit from chat attachments (ch. L1.1): the knowledge base is designed for many reference files, not for a few huge files.

Memory and scope

The Project gives your work a boundary. The instructions and files stay confined to that Project: they don’t mix with the other chats. This is a double advantage — the right context is always there, and the wrong one stays out. When you switch activities, you switch Project, and the model doesn’t drag along material that doesn’t belong.

Example: this book’s Project

This manual was written inside a Project. Its shape shows the pattern well:

Result: every writing session starts knowing what the book is like. It’s the difference between explaining the project from scratch every time and having written it once.

In practice: create your Project

  1. In Claude, open the Projects section and create a new one.
  2. Give it a name that says what it’s about.
  3. Write the instructions: the stable context, the constraints, the format you want.
  4. Upload the recurring reference files into the knowledge base.
  5. Open a chat inside the Project and work: the context is already active.
  6. When a rule recurs across several chats, move it into the Project’s instructions.

Common mistakes

Summary

  1. A Project is a workspace with instructions, reference files and linked chats, all in one place.
  2. Create it when you come back to a piece of work several times and repeat the same context.
  3. The instructions apply to every chat in the Project; the knowledge base is the material Claude consults.
  4. The scope keeps the right context in and the irrelevant out.
  5. Free up to 5 Projects; team sharing on Team/Enterprise.

Next step

In ch. L3.3 — Connectors we connect Claude to the apps you already use — Drive, Slack, task managers — so it can read your data and act within them.

Details on Projects (Free limits, team sharing) from the ledger, verified on 22/06/2026 on support.claude.com. The example of this book’s Project is real; the operational steps were not run here.

Found it useful? Get the whole manual

33 chapters in a single A5 PDF, free and no signup.

↓ Download the free PDF