01 / At a glance
Claude in one page
Strong on long documents, analysis, and writing. Personal plans train on your chats only if you allow it. Team and Enterprise do not.
| Plans | Free, Pro, Max. Team and Enterprise for firms. |
|---|---|
| Trains on your chats? | Free, Pro, Max: only if you allow it, a choice made at signup or in Privacy settings. Team and Enterprise: no, by default. |
| Your instructions live in | Settings > Your Claude > Instructions |
| Instruction limit | No limit is published. Keep it short and put the key rules first. |
| Memory | Memory from chats. Settings > Your Claude > Your memory |
| Workspaces | Projects: knowledge files, instructions, and their own memory |
| Custom assistants | Projects and Skills (Customize > Skills) |
| Turn off training | Settings > Privacy > Help improve our AI models |
The setup path
2026 change: settings moved
Since September 2026, chat and Cowork are one Claude. Settings now sit under Your Claude and Customize.
It is still rolling out. On older layouts, instructions sit under Account > Instructions for Claude.
02 / Set it up
Five steps, in order
Do them in this order. Data controls come before anything personal goes in.
claude.com/pricing
Use Team or Enterprise for firm work. Neither trains on your content by default. Audit logs and custom retention are Enterprise only.
Settings > Privacy
On Free, Pro, or Max, check Help improve our AI models and make sure it is off. If it is on, de-identified data may be kept for up to five years.
Settings > Your Claude > Instructions
Paste the instructions the master prompt writes for you. They apply to every chat.
Projects > + New Project
One project per recurring task. Add knowledge files and project instructions. Each project keeps its own memory, so client work stays separate.
Settings > Your Claude > Your memory > Manage
Read what it remembers. Pause memory keeps it. Reset memory deletes it for good. Check Customize > Connectors too.
03 / The master prompt
Let Claude set itself up
Paste this into a new chat. It interviews you, then writes your setup for you.
You are my setup chief of staff for Claude. Your job: set Claude up so it works the way I work, safely. Work in three stages. Ask one question at a time. Wait for my answer. STAGE 1. INTERVIEW (no more than 10 questions) Ask about: my role and organization, the work I want help with, the five tasks I repeat most, the tools I use every day, how I like answers written (length, tone, format), and what I must never share with an AI tool. STAGE 2. WRITE MY SETUP a) My personal instructions text. Keep it under 1,500 characters. Write it in plain sentences. Put the most important rules first. b) Three Projects I should create. For each one: a name, its purpose, its instructions, and the files I should add. c) Two Skills for my most repeated tasks. For each: a name, when to use it, and the steps it should follow. d) A checklist of the privacy settings I should confirm, with where to find each one in Claude. STAGE 3. TEST Give me three test prompts that prove the setup works, and describe what a good answer looks like for each. RULES - Do not invent features. If you are not sure a setting exists in Claude today, say so and tell me where to check. - Never ask me for passwords, account numbers, client names, or personal data. If I start to share them, stop me. - Plain language. Short sentences. Tables where they help.
04 / Govern first
Best practices
- Turn off model training before your first real prompt
- Use an approved business tier for firm work, or nothing
- Keep one workspace per recurring task
- Check every number and citation it gives you
- A person signs off before anything leaves the building
- Passwords, API keys, or access tokens
- Client names, account numbers, or holdings
- Material non-public information
- Personal data about staff or members
- Anything under a confidentiality agreement
Specific to Claude
- Use Incognito chat (the ghost icon) for one-off sensitive questions. It is not trained on and is kept 30 days by default. It does not work inside projects.
- To skip memory for a single chat, turn off Memory in the + menu before the first message.
- On Team and Enterprise, owners approve each connector and can set every tool to Always allow, Needs approval, or Blocked.
- Enterprise can scan uploaded Skills for malicious content. Turn it on before staff share Skills.
Starter instructions
Prefer to write them yourself? Fill in the brackets, then paste into Settings > Your Claude > Instructions.
About me: I am a [role] at a [type of organization]. I use you for [three main jobs]. How to answer: - Lead with the answer, then the reason. - Keep it short. Use tables for comparisons. - Say plainly when you are unsure, and how I can check. - Cite a source for every number or factual claim. Never: - Invent figures, quotes, or sources. - Give a final legal, tax, or investment decision. Flag it for a person. - Ask me for client names, account numbers, or passwords.
05 / Data and controls
What to check, and where
| Training opt-out | Settings > Privacy > Help improve our AI models (Free, Pro, Max) |
|---|---|
| Incognito chat | Ghost icon on a new chat. Not trained on. Kept 30 days by default. |
| Memory | Settings > Your Claude > Your memory. Pause or reset. |
| Export | Settings > Privacy. On Team and Enterprise, an owner must enable it. |
| Deletion | Deleted chats leave back-end systems within 30 days. |
| SSO | Team and Enterprise |
| Audit and retention | Enterprise only: audit logs, Compliance API, custom retention. |
| Data residency | Enterprise can keep processing in the US. Ask Anthropic about other regions. |
06 / Sources
Where every fact comes from
- Claude pricing and plans
https://claude.com/pricing - Is my data used for model training?
https://privacy.claude.com/en/articles/10023580-is-my-data-used-for-model-training - Anthropic: updates to consumer terms
https://www.anthropic.com/news/updates-to-our-consumer-terms - How long do you store my data?
https://privacy.claude.com/en/articles/10023548-how-long-do-you-store-my-data - Understanding Claude's personalization features
https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features - Memory in Claude
https://support.claude.com/en/articles/11817273 - Create and manage projects
https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects - What are Skills?
https://support.claude.com/en/articles/12512176-what-are-skills - Use connectors
https://support.claude.com/en/articles/11176164-use-connectors-to-extend-claude-s-capabilities - Change model improvement settings
https://privacy.claude.com/en/articles/12109829-how-do-i-change-my-model-improvement-privacy-settings - Use incognito chats
https://support.claude.com/en/articles/12260368-use-incognito-chats - Export your Claude data
https://support.claude.com/en/articles/9450526-export-your-claude-data - Set up single sign-on
https://support.claude.com/en/articles/13132885-set-up-single-sign-on-sso - Access audit logs
https://support.claude.com/en/articles/9970975-access-audit-logs - Custom data retention for Enterprise
https://support.claude.com/en/articles/10440198-configure-custom-data-retention-controls-for-enterprise-plans - Enable US-only inference
https://support.claude.com/en/articles/15422948-enable-us-only-inference-for-your-organization - Claude Cowork and chat are one Claude
https://support.claude.com/en/articles/16761823-claude-cowork-and-chat-are-one-claude
Settings checked on 29 October 2026. Platforms change their menus often. Follow the source links before relying on a path. This guide is general information. It is not legal, compliance, or investment advice.