Hasty Briefsbeta

Bilingual

The Complete Claude Setup Checklist: 72 Steps from Default to Power User

2 hours ago
  • #AI Tool Setup
  • #Claude Configuration
  • #Productivity Workflow
  • Use Claude Pro ($20/month) to unlock Projects, extended context, priority access to Opus, and Claude Code.
  • Download the desktop app for Cowork mode and local file access.
  • Set Opus 4.7 with Extended Thinking for complex reasoning tasks.
  • Turn on Computer Use for screen control (Mac-only) and connect Dispatch to your phone for cross-device context.
  • Install integrations like Claude in Excel, Wispr Flow for voice input, and Obsidian for markdown editing.
  • Create a Claude Cowork folder with ABOUT ME, OUTPUTS, and TEMPLATES subfolders for structured context.
  • Write concise about-me.md (<2,000 words) and anti-ai-writing-style.md files to personalize outputs and ban unwanted styles.
  • Set global instructions in Settings and ensure Cowork reads ABOUT ME but not OUTPUTS or TEMPLATES by default.
  • Use effective prompting: start with goals and success criteria, use AskUserQuestion, give examples, batch tasks, and open new chats for topic changes.
  • Connect tools like Gmail, Google Drive, Notion, Calendar, GitHub, Granola, and Gamma, but disable unused connectors to save tokens.
  • Build Skills via Claude's skill-creator, test rigorously, save in Capabilities, and add "do NOT use for" clauses to prevent false triggers.
  • Organize work with Projects (one per deliverable), upload gold-standard examples, keep instructions under 8 lines, and use scheduled tasks.
  • For development, use Claude Code with a CLAUDE.md file per project, enable autoMemoryEnabled, and understand the memory hierarchy.
  • Use Claude Design for web pages and Claude Code for full applications, connecting GitHub for deployment.
  • Enable Web Search, use Research mode for deep reports, prioritize recent sources (2025-2026), and save findings to files.
  • Manage tokens efficiently: convert files to markdown, restart conversations every 20 messages, switch models based on task complexity, and spread sessions to avoid rate limits.