Check your Claude settings
Last checked . Checked against Understanding claude s personalization features, Use claude s chat search and memory to build on previous context, How do i change my model improvement privacy settings, Create and edit files with claude, Use connectors to extend claude s capabilities, Manage plugins skills and connectors from one plugins page, What are projects, Change the model effort and thinking settings, Schedule recurring tasks in claude cowork.
In short. Open Settings and set standing instructions, memory, and data controls. Then turn on file creation, add one read-only connector, and make a project for each job. Plan for about 15 minutes, once.
Before you start
Finish Set up Claude first. Menu names change often. Each step below says what to look for if your screen differs. On a work account, an owner may lock some of these, so ask IT when a switch is grey.
Set these once
Step 1. Write your standing instructions
Click your initials in the lower left, open Settings, then Your Claude. On older screens, open Settings, then Account. Find Instructions for Claude and type your text. Claude reads it in every chat, so you stop retyping who you are. Read What is a system prompt for why this works, and How to stack context for AI for where it fits.
Here is a copy-ready start for a buyer:
I am a buyer at Mesquite Ridge Manufacturing. I compare supplier quotes for machined parts.
Use US units and US dollars. Keep part numbers exactly as written, including dashes.
If a quote does not state something, write "not stated". Do not guess.
Write drafts only. I send every email myself.
Use short sentences and plain words.
Keep it short. Change it when a rule keeps being wrong.
Step 2. Choose your memory settings
In Settings, then Your Claude (older screens: Settings, then Memory), look at two switches.
- Search and reference chats lets Claude look through your past chats. It is on by default on paid plans.
- Generate memory from chats lets Claude save notes about you and carry them into new chats. If you turn it off, you can choose Pause memory (keeps the notes, stops using them) or Reset memory (deletes all notes for good).
For quote work, many people switch memory off, so old quotes do not leak into new answers. Read Does AI remember past chats before you decide. On Team and Enterprise plans, an owner controls memory.
Step 3. Check the data controls
Open Settings, then Privacy. Find the switch that lets Anthropic use your chats to improve its models. This applies to Free, Pro, and Max accounts. Turn it off if you will use company or customer files. Turning it off stops use of new chats but does not undo past use. Work plans (Team, Enterprise) are handled under separate commercial terms, so ask IT. Read Does AI learn from my data and the safe use notes.
Step 4. Confirm file creation and network access
Open Settings, then Capabilities. Check that Create and edit files is on (older name: Code execution and file creation). This is also where Allow network egress sits. Network access lets Claude's file tool reach the internet. For quote files, turn it off unless a guide tells you otherwise. On Team and Enterprise, an owner sets this, and All domains is the riskiest choice.
There is no separate artifacts switch. Artifacts work when file creation is on.
Step 5. Add one connector, read-only first
Open Plugins in the sidebar (older: Customize), then Connectors, and click +. Pick one, such as your email or file storage, and click Connect. Sign in when asked. Start with one. On Team and Enterprise, an owner must allow it first, and owners can set a tool to Always allow, Needs approval, or Blocked. Ask for read-only so Claude can search email but not send. To remove one, go back to Connectors and choose to disconnect it. Removing a plugin does not disconnect its connector, so disconnect separately. Read How AI connects to your systems and What is MCP.
Step 6. Make a project for each job
Click Projects in the sidebar, then create a new project. Name it for the job, such as "Supplier quotes". Add its files and its own instructions. Chats inside it stay together. Projects work on all plans, including free (up to five). Sharing is for Team and Enterprise. Read What is an AI project.
Step 7. Look at skills and plugins
Open Plugins in the sidebar. Skills, connectors, and plugins now share this page. Public ones carry a Verified or Community label. Only add what you trust. To add your own skill, follow Upload a Claude skill.
Step 8. Know the model and effort menu
Click the model name next to the send button. Pick a model, then Effort. Higher effort is slower and uses your limit faster. Use Low or Medium for simple jobs and High when accuracy matters. Each model marks its own recommended level as Default.
Step 9. Set up a scheduled task
A scheduled task is a prompt that Claude runs for you at a set time, such as every Monday morning.
Today it lives on the Scheduled page in the left sidebar (in the newer Claude screen, you may only see it after you ask for one in a chat). To make one, open any chat and describe the job and how often it should run. Answer Claude's questions. Then check the name, schedule, and instructions, and click Schedule. Or click Scheduled, then New task, then Set up manually, and fill in the name, prompt, approval mode, and how often.
Here is a sample for a buyer:
Every Monday at 7 am, list open POs past due from the attached export and draft one chase email per supplier. Do not send anything.
To see your tasks, open Scheduled. Click a task to see past and upcoming runs, edit it, Pause or resume it, run it now, or delete it. Paid plans have it (Pro, Max, Ultra, Team, and Enterprise). On Team and Enterprise, an owner may need to turn it on. If you pick a folder, the task may run only while your computer is awake and Claude Desktop is open. If a connector is set to ask first, the task may wait for you or finish without it.
Safety habit. Start with drafts only. Say "do not send" in the prompt. Read the first few runs before you trust it. Read How to schedule AI tasks and What triggers an AI agent before you rely on one.
Check it worked
Open a new chat and ask: "Who am I, and what rules should you follow?" Claude should repeat your role and your "say not stated" rule. Then ask it to list your connectors. Only the one you added should appear.
If something goes wrong
- Claude ignores your instructions. Open a new chat, because an old chat may hold different rules. Shorten the text to the most important lines.
- You cannot find Your Claude. Try Settings, then Account or Memory. Screens differ by app version.
- A switch is grey. An owner has locked it on a Team or Enterprise plan.
- A connector asks for too much access. Cancel and pick a narrower option, or skip it.