Skip to content

Scenario Hands-On: Choose Your Track ​

💡 Goal: The 11 lessons are split into three tracks by audience. Pick one and finish it in about 36–57 minutes—each lesson produces a finished result you can verify on the spot: an AI assistant at your beck and call in your IM, a morning briefing that arrives on time every day, or a Claw that writes code under approval.

The first three stages took care of "installed," "chatting smoothly," and "can automate": the Gateway is running, models are connected, and the three components—scheduled tasks, browser, and MCP—are all in your hands.

This stage answers a practical question—what can OpenClaw hands-on practice help you do? The 11 scenario lessons are split into three tracks by audience: office workers plug the AI assistant into WeCom, Feishu, or WeChat; creators set up morning briefings, watch web pages, and turn images into speech; programmers let Claw write code, stand GitHub watch, and deploy to servers. Just pick the track you'll actually use—you don't need to learn them all.

🎯 Stage Goals ​

  • Hook Claw into at least one chat channel you genuinely use: openclaw channels status --probe shows the channel online, and you receive its replies in the chat window
  • Get at least one unattended automation up and running: a morning briefing appears in your IM on time every day, or you get notified the moment a watched web page changes
  • If you take the programmer track, also let Claw write code and run commands under approval—changes are presented as patches and only land once you nod

⏳ Estimated Time ​

  • The 11 lessons total about 150 minutes (105–195), but the three tracks are independent of each other—just pick one and finish it:
    • Track A (Office): 4 lessons, about 57 minutes (40–74)
    • Track B (Creative): 3 lessons, about 36 minutes (25–47)
    • Track C (Programmer): 4 lessons, about 57 minutes (40–74)
  • Each lesson stands on its own and can be read individually; the mechanisms involved all come from the first three stages, so once you've learned them you can assemble directly

🚀 Before You Start ​

  • You've completed Daily Use: the Gateway is running and at least one model is connected (/model status can report the current model)
  • Before taking Track B, catch up on two lessons first: the morning briefing needs 3.1 Scheduled Tasks, and web page monitoring needs 3.2 Browser Control—the scenario lessons only handle assembly and won't re-explain the mechanisms
  • If you want to connect domestic IMs: the QQ bot requires creating a bot on the q.qq.com open platform first to get an AppID/AppSecret; preparation items for WeCom, Feishu, and Yuanbao are at the start of each lesson
  • A cup of water ☕—the endpoint of all three tracks is "a finished result running in a real environment"

📚 Choose Your Track ​

Before choosing a track, let's answer a frequently asked question: Which chat apps does OpenClaw support? Here are the results of verifying each mainstream domestic IM one by one (based on the official channels documentation, 2026-11 verified):

ChannelPlugin (maintainer)Supported scope
Feishu@openclaw/feishu(OpenClaw official)Direct chat + group chat, streaming card replies, with Feishu Docs/wiki/cloud drive tools
WeCom@wecom/wecom-openclaw-plugin(Tencent WeCom team)Bot/Agent dual mode, direct chat + group chat, streaming replies, proactive push
WeChat@tencent-weixin/openclaw-weixin(Tencent WeChat team)Direct chat only(including media), group chat not supported
QQ Bot@tencent-connect/openclaw-qqbot(official plugin)C2C direct chat + group @ mentions, supports images/voice/video/files
Yuanbaoopenclaw-plugin-yuanbao(Tencent Yuanbao team)Direct chat + group chat (WebSocket)
DingTalk@dingtalk-real-ai/dingtalk-connector (DingTalk team, not part of OpenClaw core)Direct chat + group chat, AI Card streaming replies, with tools for DingTalk Docs, calendars, to-dos, and more; see A5

⚠️ Before choosing a channel, three honest disclosures

  • DingTalk: OpenClaw core does not support it; an external plugin is required. The official OpenClaw documentation has no DingTalk channel (re-checked on 2026-11-20); the DingTalk team maintains a plugin on GitHub self-described as "officially produced by DingTalk" called dingtalk-openclaw-connector—see A5 DingTalk integration for how to connect it. The plugin's README suggests using it as a personal assistant and avoiding direct deployment in enterprise production environments.
  • WeChat: direct chat only. The plugin's capability statement says direct chats only—it can chat 1-on-1 and send/receive media, but it cannot enter WeChat group chats.
  • Yuanbao: details not verified against the core. The channel entry comes from the plugin maintainer's (Tencent Yuanbao team) documentation; OpenClaw has not verified it against the core source code, so when configuring, defer to the A3 page.

As for international channels, OpenClaw also supports Telegram, Slack, Discord, WhatsApp, iMessage, Matrix, Microsoft Teams, Signal, Google Chat, X, and 20+ other platforms, all connected with the same command: openclaw channels add. The official recommendation is to connect Telegram first—just one bot token, the fastest to get running:

bash
openclaw channels add --channel telegram --token <bot-token>

The three tracks are as follows—each picks their own:

Track A: Office Workers—Plug the AI Assistant into the IM You Already Use ​

This is for you if: your daily main battlefield is WeCom, Feishu, DingTalk, WeChat, or QQ, and you want an AI assistant to appear directly in those chat windows instead of being yet another webpage you have to open separately.

CourseWhat you will buildDuration
A1 WeCom IntegrationAI assistant in WeCom: Bot/Agent dual modes, streaming replies, proactive push15 minutes
A2 Feishu IntegrationFeishu bot: chats in both DMs and groups, and can operate Feishu docs/wiki/cloud drive15 minutes
A3 WeChat, QQ, and YuanbaoThree Tencent channels: WeChat DM assistant, QQ group bot, Yuanbao15 minutes
A4 Email AssistantIMAP inbox on duty: allowed incoming mail is handled in isolated sessions automatically (no emails sent)12 minutes
A5 DingTalk integrationDingTalk team's external plugin: install with one command, create a bot by scanning a QR code—

The first 4 lessons of Track A take about 57 minutes (40–74); A5 DingTalk was added in 2026-11 and is not counted. Connect the IM you use most first; A4 does not depend on any IM and can be inserted at any time. The Feishu/WeCom/Yuanbao plugins have minimum version requirements for OpenClaw (2026.5.29+ / 2026.3.22+ / 2026.4.10+), and this tutorial's baseline v2026.10.9 meets them all.

Track B: Creators — Information Streams and Content Production Automation ​

This fits you if: you run social media accounts, write for a WeChat official account, or simply consume too much information — you want AI to fetch news on a schedule, watch webpages, add images, and read articles aloud for you.

CourseWhat you will buildDuration
B1 Daily Morning BriefFetches news at a fixed time every day, summarizes it into a morning brief, and sends it to your IM12 minutes
B2 Webpage MonitorWatches target pages and notifies you the moment they change (price monitoring, update alerts)12 minutes
B3 Images and VoiceAdds images to your content and converts text to speech12 minutes

Track B has 3 courses, about 36 minutes in total (25–47). B1 and B2 are combined applications of 3.1 scheduled tasks and 3.2 browser control; to deliver the morning brief into an IM, borrow a channel from Track A — skipping it is fine too, the tasks still run (for how to view run records, see 3.1).

Track C: Programmers — Coding, On-Call Duty, and Deployment ​

This fits you if: you write code and want AI to modify code and run commands under your approval, take your GitHub on-call shifts, and finally deploy to a server running 7×24.

CourseWhat you will buildDuration
C1 Let Claw Write CodeRuns commands via exec, presents changes as patches, applies them only after approval15 minutes
C2 GitHub On-CallWebhook wakes the agent: PR summaries, automatic issue handling12 minutes
C3 Multi-Agent CollaborationOne Gateway runs multiple isolated expert agents, bound by channel15 minutes
C4 Server DeploymentRuns 7×24 on Docker/VPS, usable even when you are away from your computer15 minutes

Track C has 4 courses, about 57 minutes in total (40–74). Dependencies: C2 builds on C1; C4 stands on its own, with alternative installation methods (desktop app, Docker, and npm) as the only prerequisite — if you want to get the service onto a server first, you can jump straight there.

🎒 What You Will Gain ​

  • An AI assistant living in your IM: on call anytime in WeCom, Feishu, WeChat, and QQ, no longer confined to a browser tab
  • A self-running information stream: the morning brief arrives on time, watched webpages alert on any change — you just check the results
  • (For programmers) A complete pipeline from coding to going live: exec and patch approval → GitHub on-call → multi-agent division of labor → server 7×24
  • An extensible map of OpenClaw scenarios: every finished product is a combination of parts from the first three stages; with the assembly approach in hand, you can build new scenarios yourself

After completing any one of the three routes, your hands-on OpenClaw experience will have its first finished product — it is not just a chat tool, it can also do work for you. To dig deeper into Skills and security hardening, you can continue anytime with the advanced handbook.


Ready? Start Lesson 1: WeCom Integration → (creators can head straight to B1 Daily Morning Brief; programmers can start from C1 Let Claw Write Code.)

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.