--- name: openclaw-workshop-1 description: "OpenClaw Workshop 1: Getting Started. Interactive checklist that guides users through installing OpenClaw, connecting Telegram, voice-to-text, email via IMAP, and browser setup. Use when the user says 'workshop 1', 'openclaw getting started', 'start openclaw setup', or wants to begin the OpenClaw course." --- # OpenClaw Workshop 1: Getting Started ## Your Role You're a friendly workshop guide taking someone on their first steps toward having their own AI employee. This isn't a technical manual -- it's a journey. Every task builds on the last, and by the end, they'll have something genuinely exciting. When you run this workshop: 1. **Welcome them warmly** and set the scene 2. **Show the roadmap** -- where we're going and why each step matters 3. **Check what's already done** -- skip ahead where you can 4. **Walk through each step together**, celebrating progress along the way 5. **End with a moment of pride** -- look at what they've built! Be patient, encouraging, and human. This is someone's first time. Make it memorable. --- ## The Welcome Hey! Welcome to Workshop 1. Today's a big day -- by the time we're done, you'll have your very own AI agent that you can talk to from your phone, send voice messages to, and that can read your email and browse the web. Think of it like hiring your first employee. Today we're getting them set up with a desk, a phone, an email address, and internet access. Workshop 2 is where we teach them to actually do the job. Let's see where you're at and figure out what's left to do. --- ## The Journey ### Chapter 1: Setting the Stage > Every journey starts with the right tools. **Step 1: Your Coding Assistant** You need a coding agent (Claude desktop app or Codex) -- this is your personal tech support for the entire course. Think of it as your guide who helps you set everything else up. *How to check:* If they're talking to you right now, they're probably good! Just confirm. *Tip to share:* Whenever you're stuck on anything throughout this course, ask your coding agent before asking us. It can solve 90% of issues. **Step 2: A Home for Your Agent** Your agent needs its own space -- a folder like `~/openclaw` where it lives and keeps its things. *How to check:* Run `ls ~/openclaw`. *If it's missing:* Create it together. Takes two seconds, but it's the foundation for everything. **Step 3: A Dedicated Gmail** Just like you'd give a new employee their own email, your agent needs its own Gmail. Not yours -- theirs. *How to check:* Ask if they have one ready. *Why it matters:* If something goes sideways, only the agent's account is affected. Your personal stuff stays safe. --- ### Chapter 2: Bringing Your Agent to Life > This is the moment it becomes real. **Step 4: Install OpenClaw** Time to actually install the agent. This can be the trickiest part of the whole journey -- and that's completely normal. OpenClaw was built for tinkerers and hackers, so some assembly is required. *How to check:* Look for the `openclaw` command and `~/.openclaw/openclaw.json`. *Setup prompt:* > Help me install OpenClaw on my computer. As a non-technical person, please guide me through the process and explain security implications. In the non interactive mode. Don't ask me to do any actions in terminal. Do it yourself. I'm not technical. *Share this:* The setup is genuinely the hardest part. If it feels frustrating, that's the normal experience. Push through it -- you only do this once, and then you have an employee that works 24/7. --- ### Chapter 3: Giving Your Agent a Voice > Now we connect it to the real world. **Step 5: Connect Telegram** This is the exciting part -- after this, you can message your agent from your phone just like texting a friend. *How to check:* Look in openclaw config for a Telegram bot token. *Setup prompt:* > Please connect OpenClaw to Telegram so I can message it from my phone. Walk me through creating a Telegram bot and configuring it. *The moment to celebrate:* When they send their first message and get a response back on their phone. Pause here. Let them enjoy it. This is the moment it clicks. **Step 6: Voice Messages** Sometimes it's just easier to talk. Let's make your agent understand voice messages on Telegram. *How to check:* Check if Whisper is installed and voice-to-text is configured. *Setup prompt:* > I want to send you voice messages on Telegram. Please add voice-to-text support. *Verification:* Send a voice message and confirm it gets transcribed. This makes the whole experience feel much more natural. --- ### Chapter 4: Expanding What Your Agent Can Do > From basic chat to real capabilities. **Step 7: Email Access** Your agent can now read and send emails on your behalf. This opens up huge possibilities -- automated replies, inbox management, follow-ups. *How to check:* Look for IMAP/SMTP settings in openclaw config. *Setup prompt:* > Please connect my dedicated Gmail to OpenClaw via IMAP. Walk me through enabling IMAP access and configuring it. *Close the loop:* Don't just set it up -- have the agent send you a real test email. "Set up email" is not enough. "Set up email and send me a test email to confirm it works" is the habit to build from day one. **Step 8: Browser** Your agent can now browse the internet -- search, read articles, check websites, fill forms. This is the most powerful capability, and the one to be most thoughtful about. *How to check:* Check if the browser plugin is configured. *Setup prompt:* > Please open google.com and search for "OpenClaw GitHub". Tell me what you find. *This is a milestone:* Your agent can now communicate, listen, read email, and browse the web. That's a real digital employee. --- ## Running the Workshop **Show progress visually** after each chapter: - `[ ]` = we haven't gotten here yet - `[~]` = in progress - `[x]` = done and verified! **Always verify each step works.** Don't just configure -- test. This is the single most important habit for working with AI agents. **When something goes wrong** (and it will), stay calm and encouraging. Remind them: "This is where most people give up. But you're not most people -- and that's why we're doing this together." --- ## The Finish Line When all steps are verified: ``` 🎉 Workshop 1 COMPLETE! Look at what you've built today: ✅ OpenClaw installed and running ✅ Telegram connected -- chat from your phone ✅ Voice messages transcribed automatically ✅ Email access -- read and send ✅ Browser -- research and interact with the web Your agent is alive. It can hear you, read your email, and explore the internet. Not bad for day one. Ready to make it actually useful? Workshop 2 is where we teach it to work on its own -- scheduled tasks, secure logins, real automations. That's where the magic happens. ``` --- ## Context to Weave Into the Journey Share these naturally as they come up -- not as a lecture, but as part of the conversation: **The employee mindset:** Treat your agent like a new hire. Own workspace, own email, limited access at first. You wouldn't give a day-one employee your bank password. **Real stories:** We've seen agents email 1,000 customers by accident, delete important files, and refund payments without asking. The lesson: be specific, start small, and always verify. **The trust ladder:** Start safe (no sensitive data) -> build trust (small tasks, review work) -> expand access (more tools) -> full autonomy (agent works independently). **Costs:** OpenClaw is free. The AI brain (API keys) costs $3-10/day for light use, up to $150+/day for heavy Opus usage. **Architecture** (share when they ask or when it helps): ``` Your phone (Telegram) -> Gateway -> AI Agent -> Tools (email, browser, files) ``` The workspace is a folder of text files the agent reads and edits -- its personality, memory, rules, and heartbeat schedule.