Skip to content

Daily use: working efficiently with your Claw ​

💡 Goal: 3 lessons, about 26 minutes, to master everyday OpenClaw usage—chat with Claw in the browser's Control UI, use slash commands to switch models and start new sessions, and get Claw to write your preferences into memory with a single sentence.

Once it's installed, "how do I use it smoothly" is the next question. Everyday use of OpenClaw boils down to three things: where to chat with it (the browser Control UI, the terminal TUI, chat channels), how to direct it (slash commands like /model and /think), and how to make it remember you (a memory system backed by visible Markdown files). This stage uses 3 lessons to cover each of the three thoroughly—state, memory, and credentials all stay on your own hardware.

🎯 Stage goals ​

  • Open Control UI anytime: openclaw dashboard one command, then manage sessions, review approvals, browse files, and change settings in the browser (default on this machine: http://127.0.0.1:18789/)
  • Direct Claw with slash commands:/model switch models, /think adjust thinking, /new start new sessions, /skill <name> tweak skills—one line in the chat box does it, no need to touch config files
  • Make Claw remember you: a single "Remember that I prefer TypeScript" lands in ~/.openclaw/workspace's Markdown memory file and can be recalled across sessions within the same agent's direct chats (enabled by default for personal installs)

⏳ Estimated time ​

  • 3 lessons totaling about 26 minutes; with buffer time, about 18–34 minutes
  • Each lesson stands on its own and can be read individually; 2.1 Control UI is the workbench for the last two lessons, so starting there is recommended

🚀 Before you start ​

  • completed the installation:5 minutes to get running: the Gateway runs, and onboarding has connected at least one model (starting from scratch? See the quick start guide first)
  • remember that model references are written as provider/model (e.g., deepseek/deepseek-v4-flash)—if it's fuzzy, go back to the model overview for an 8-minute refresher
  • no desire to touch the terminal? That's fine: most operations in this stage happen in the browser
  • a cup of water ☕—all three lessons focus on building a feel for it, and every command is short

📚 Lessons in this stage ​

Daily use answers three questions, each with its own lesson:

Everyday questionThis stage's answerLesson
Where is it most convenient to chat with Claw?Control UI web version: a browser console centered on conversation, with files, approvals, and settings right alongside2.1 Control UI
How do I quickly switch models, adjust thinking, or start a new session?Slash commands: /model, /think, /new…type one line in the chat box and it takes effect2.2 Chat commands
How do I get Claw to remember my preferences and project context?Memory system: plain Markdown files + a SQLite index; one sentence and it's written2.3 Memory system

Lesson schedule:

LessonContentDuration
2.1 Control UI: the console in your browserOpening http://127.0.0.1:18789/, connection and pairing, sessions and sidebar, chat and approval panels, the settings page8 minutes
2.2 Chat commands crash course: /model, /think, and slash commands/model family for switching models, /think to control thinking, /new and /reset for sessions, /skill as the universal entry point, skills as commands8 minutes
2.3 Memory system: making Claw remember youFour memory files and three tools, cross-session recall, Chinese trigram retrieval, importing memories from Claude Code / Codex / Hermes10 minutes
A quick look first: three kinds of entry points for chatting with Claw
  • Browser Control UI:openclaw dashboard to open, default on this machine http://127.0.0.1:18789/—sessions, approvals, files, and settings all live here; 2.1 covers it for a whole lesson
  • Terminal TUI: type openclaw in the terminal to enter—chat without opening a browser
  • Chat channels: Hook Claw into the IM apps you already use, such as Feishu, WeCom, and Telegram—Hands-on Scenarios walks you through configuring each one

⚠️ Don't enable debugging commands in group chats

/reasoning, /verbose, and /trace bring internal reasoning, tool parameters, plugin diagnostics, and other content not suitable for public viewing into the channel. Keep them off in group chats, and only enable them in trusted private chats or controlled rooms.

🎒 What you'll gain ​

  • A smooth daily control panel: start sessions, review approvals, and change settings in the browser—openclaw dashboard takes you back to the scene anytime
  • A feel for a set of high-frequency commands: switch models, adjust reasoning, start new sessions, and tune skills—all done in one line from the chat box
  • A Claw that remembers you: preferences are written into visible Markdown files, recallable across sessions, and you can even import old memories from Claude Code / Codex / Hermes
  • The road map ahead: Efficient Workflows covers scheduled tasks, browser control, and MCP; Hands-on Scenarios hooks Claw into WeCom and Feishu

Once you finish these three lessons, your everyday ways of using OpenClaw will be complete—chat in the browser, tune from the command line, accumulate in memory. Next you can go to Efficient Workflows: scheduled tasks, browser control, and the MCP add-on toolbox; if you'd rather see the finished product right away, the daily morning briefing in Hands-on Scenarios is built from scheduled tasks and web tools.


Ready? Start the first lesson: Control UI →

gaodong.aiAboutPrivacyContact

Independent tutorials, not affiliated with any product mentioned. Product names and trademarks belong to their owners. Check official pages for prices and policies.