Quick Answer
Stop re-explaining the same context every time you use Claude.
Primary-source note: Anthropic describes Projects as self-contained workspaces with their own chats and knowledge bases.
Key facts
What to know first
Use Projects for durable context
A Project should have a purpose narrow enough that every file and chat belongs. Good examples: “Q3 board reporting,” “Spanish 201 coursework,” “Acme renewal,” “2026 tax prep,” or “home renovation.” Bad examples: “all work,” “personal,” or “random notes.”
Add stable context to project knowledge: policies, templates, examples, background docs, glossary, style guide, and decision history. Keep volatile or sensitive material out unless it is necessary and allowed.
- Name the project by outcome, not category.
- Add only reference files that will matter repeatedly.
- Write project instructions: role, goals, tone, exclusions, and output standards.
- Review project knowledge monthly and remove stale files.
Source check: What are projects? from Claude Help Center.
Memory now follows you between chat and Cowork
Anthropic says the memory you use in chat is the same memory Cowork uses. Explain once in chat how your team defines a metric, and a deck Cowork builds later uses it without a rebrief; what comes up in a Cowork task carries back to chat. Memory also updates as you talk rather than by summarizing a conversation after it ends, so a deadline you mention mid-chat is already known next time.
Everything Claude remembers is a short file under Topics in Memory settings. You can read, edit, or delete any of it, and Anthropic points out the leverage in that: fix your company’s old name in one file and every later conversation gets it right. You can pause or reset memory at any time.
Defaults differ by plan. Memory is on by default on Free, Pro, and Max across web, desktop, and mobile. On Team and Enterprise, admins control availability and memory stays off for individual users until they turn it on.
- Open Settings, then Memory, and read the Topics list before you trust it.
- Delete anything stale; a wrong memory is worse than a missing one.
- Decide deliberately whether to enable sensitive topics.
- Keep facts that must be exact in project knowledge, not memory.
Source check: Claude's memory works everywhere, and you decide what's in it from Claude by Anthropic.
Decide what memory is allowed to hold
By default Claude does not store topics it treats as personal or sensitive — health, race, ethnicity, religious beliefs, politics, gender identity, and similar. You can turn "include sensitive topics in memory" on if re-explaining them is the actual friction, for example a food allergy that should shape every meal-planning answer. Anthropic says it notifies you each time it saves one, and that the setting is forward-looking only: nothing from before you enabled it is saved retroactively.
Some categories stay out regardless. Anthropic lists sensitive identification numbers such as SSN or government IDs, criminal history, immigration status, and anything violating its Acceptable Use Policy; Claude tells you when it cannot store something. Memory is still not the right home for facts that must be exact — put those in project knowledge or the source document, where you can see and version them.
Source check: Claude's memory works everywhere, and you decide what's in it from Claude by Anthropic.
Treat skills as repeatable procedures
Anthropic’s 2026 styles-to-skills migration shows a broader product direction: reusable instructions are becoming more important. For everyday work, think of a skill as a procedure: intake questions, source requirements, output template, review steps, and failure conditions.
A good reusable instruction does not say “write well.” It says how to process the task: inspect sources, ask missing questions, draft in a named format, flag unsupported claims, and produce a final checklist.
Source check: Styles are moving to skills from Claude Help Center.
Keep project knowledge clean
The failure mode of Projects is context rot. Old files, superseded drafts, and broad instructions can pull Claude toward the wrong answer. Add a simple freshness note to every project: which files are canonical, which are examples, and which are archived.
For teams, decide who owns updates. Claude can help produce a “project hygiene” checklist, but a person should own what stays in the knowledge base.
Prompts to Copy
Design a Claude Project for [recurring work]. Specify project purpose, files to add, files to exclude, project instructions, and monthly maintenance checklist.
Review this project context. Identify stale, conflicting, missing, or overly broad instructions before we use it for new work.
Turn this repeated workflow into a reusable instruction: intake questions, source requirements, process steps, output format, and review checklist.
Cite this page
Claude Helps, "How to Use Claude Projects, Memory, and Skills", https://claudehelps.com/projects-memory (updated 2026-09-19).
Primary Sources
- What are projects?Claude Help Center
- Claude's memory works everywhere, and you decide what's in itClaude by Anthropic
- Use Claude's chat search and memoryClaude Help Center
- Styles are moving to skillsClaude Help Center
- How do usage and length limits work?Claude Help Center
FAQ
What are Claude Projects best for?
Projects are best for recurring work where the same context, files, instructions, or chat history should stay together.
Is memory the same as project knowledge?
No. Memory is a list of topic files Claude maintains about you across surfaces. Project knowledge is deliberate context you add to one workspace and can version yourself.
Does Claude remember things between chat and Cowork?
Yes. Anthropic says chat memory and Cowork memory are the same memory, so context built in one is available in the other.
Can I see and delete what Claude remembers?
Yes. Memory settings list everything as short files under Topics, where you can read, edit, or delete each one, and pause or reset memory entirely.
Is memory on by default at work?
Not necessarily. It is on by default for Free, Pro, and Max. On Team and Enterprise, admins control availability and it stays off for each user until they turn it on.
How many projects can free users create?
Anthropic’s Projects help page says free users can create a maximum of five projects.