@undefinedKi - Claude Setup Guide Anthropic

Open original

@undefinedKi on X This is a twelve-step walkthrough for setting up Claude once so it remembers your context, reuses your methods, and works across your tools without re-explaining everything each session. It draws on Anthropic's release notes and says which features live in the chat app versus Cowork.

  • Start with memory and projects, since skills, connectors, and scheduled tasks all read from them. Projects keep separate instructions, files, and memory per recurring job.
  • Save repeated methods as skills, written after you've done the task manually twice, and include when to use them, steps, output format, and what never to do.
  • Connect tools like Gmail, Google Calendar, Drive, Notion, Slack, and GitHub in one sitting, and run scheduled tasks read-only at first, checking the first run before trusting it.

Claude keeps no memory between chats by default, so the article's core argument is that you can set things up once and stop re-explaining yourself every session. The setup has three layers. Memory, under Settings, is where you turn on chat referencing and memory generation, then read and correct the four auto-generated sections (Preferences, Profile, Areas, People). Projects hold context for recurring jobs: each gets its own instructions, uploaded reference files, and memory, so a newsletter project doesn't bleed into client work. Skills, under Settings, capture repeatable methods, and the author's advice is to do a task manually twice, get it right, then have Claude write the skill from what you actually did, since skills invented from imagination tend to be guesses about your process. Each skill should cover when to use it, required inputs, steps, output format, and what it must never do. The rest of the list extends the same idea outward. Connectors (Gmail, Google Calendar, Drive, Notion, Slack, GitHub, Linear or Jira) let Claude pull context without pasting, and the author recommends connecting them all in one sitting and then testing with a small, verifiable question, since a connector can authorize but be scoped wrong without any visible error. Scheduled tasks run cloud-side on a timer, useful for a morning inbox digest or a Friday summary of repo changes; the advice is to keep early tasks read-only and to check the first run rather than trusting it. On the output side, highlighting a section and asking for a change rewrites only that part, and asking for actual files (Excel with live formulas, PowerPoint, Word, PDF) produces work you can use instead of answers you still have to convert. The Excel and PowerPoint add-ins let Claude edit an open document directly, sharing context across both. The practical caveat running through the piece is that everything depends on checking the first output carefully. Formulas, page counts, and scoping errors are easy to miss in a chat answer and expensive once a bad scheduled report has been arriving for a month.

Ask about this article