hello-claw: a Datawhale tutorial track for adopting and rebuilding OpenClaw
哈喽！龙虾 🙋♀️ Adopt from scratch and build your first claw 🦞 来领养你的第一只龙虾！
At a glance
- What is it?
- hello-claw is a Chinese-language tutorial repository that treats OpenClaw as something you adopt before you modify, splitting the material into a usage track, a Skills scenario track, and a source-level development track. The judgement below is about what that structure is good for and where it stops being enough.
- Who is it for?
- Adopt hello-claw if you want a structured Chinese-language path into OpenClaw and you are willing to follow the repository's own chapter order rather than skim for commands. Skip it if you need an English-first reference, an API-level spec, or a turnkey deployment.
- Can I use it commercially?
- Not without permission. GitHub finds no licence file in the repository, and without a licence all rights are reserved by default: you may read the code but not reuse it. Check the README, or ask the authors, before using it.
- Is it still maintained?
- Yes. The repository last received commits 27 days ago.
- What is it written in?
- Mainly JavaScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
The gap hello-claw is trying to fill
OpenClaw is a command-line AI assistant system with channels, models, agents, scheduled tasks, a gateway and a Skills format. That is a lot of surface area for someone who just wants an assistant answering messages in Feishu or Telegram. The repository's own framing is adoption first: the README describes a path where a beginner picks a Claw, sends it to school, or writes one that is not bound by someone else's definition. The stated audiences are four: people with no programming background who want an always-available assistant, users who want to control the AI remotely through QQ, Feishu or Telegram, technically curious readers interested in the Skills system, and developers who want to understand the agent architecture and build their own version. The learning advice in the README is explicit about routing. Beginners start with the adoption track and get installation plus basic automation working. Readers who want an end-to-end scenario go straight to the university section and pick five to ten Skills. Developers go to the build track and take apart the underlying implementation. That routing advice is the most useful sentence in the README, because the repository is large enough that reading it linearly is a poor use of a weekend.
Three tracks, eleven chapters each, and a lot of appendices
The material is organised as three modules. The adoption track runs eleven chapters plus seven appendices, grouped into installation (chapters 1 to 3), core configuration (4 to 6), extended operations (7 to 9), and security plus clients (10 to 11). Installation covers an AutoClaw desktop client for a zero-setup first run, a manual path through Node.js, npm install and an onboard configuration wizard, and a third chapter on the initial configuration wizard including macOS guidance, custom providers, and reconfiguration. Core configuration handles chat platform integration (Feishu is the worked example, with pairing and group chat), model management (multiple providers, API key rotation, failover), and agent management (multiple agents, workspaces, heartbeats, binding rules). Extended operations covers tools and scheduled tasks with cron, at and every schedules, gateway operations including hot reload and sandbox policy, and remote access through SSH tunnels and Tailscale. The final block is security and clients: threat model, VM isolation, trust boundaries, MITRE ATLAS, supply chain, then dashboard, WebChat, Control UI, TUI and third-party clients. The university track is a menu rather than a curriculum. It lists scenarios across personal productivity (email assistant, local health assistant, morning briefing, calendar), programming (vibe coding, CI/CD assistant, documentation generation), content creation, business and sales (CRM assistant, meeting scheduling), multi-agent collaboration, and a miscellaneous group including security checklists, paper push notifications, smart home control, finance analysis and training assistance. The build track is eleven chapters that first dissect OpenClaw's source and alternatives, then move into Skills, channels and full customisation. A beginner reading the build track before the adoption track will be reading about a system they have never configured.
What the chapters actually contain, based on the changelog
The repository keeps a dated activity log, and it is the most concrete evidence of what is inside each chapter. The 2026-03-08 entry records completion of adoption chapters 1 to 11, listing the topics as installation (AutoClaw plus manual plus wizard), core configuration (chat platform, model, agent), extended operations (tools and scheduled tasks, gateway, remote access) and security plus clients (protection, web interface). The 2026-03-12 entry records build chapters 1 to 10 covering prompt system, tool system, message loop and multi-channel integration, then alternative approaches (lightweight, hardened, hardware) and a closing retrospective. A separate entry records build chapter 13 on Skill file structure, frontmatter, asynchronous handling and debugging. Note the gap: chapter 11 and 12 of the build track are not mentioned in any changelog entry in the supplied material, and the README table for the build track is not reproduced in full here, so the exact chapter titles beyond those listed cannot be confirmed from this material. The university track was rewritten for beginners on 2026-03-25, adding eleven scenario articles and reorganising them by the README's own categories. Treat the changelog as the authoritative map of what is finished, and treat the chapter tables as the intended shape.
Version tracking is the repository's real maintenance mechanism
The changelog entries pair tutorial updates with upstream OpenClaw releases, which tells you how the maintainers handle drift. The 2026-03-25 entry describes OpenClaw v2026.3.24 adding Gateway OpenAI-compatible endpoints at /v1/models and /v1/embeddings, a Microsoft Teams SDK integration with streaming replies and welcome cards, one-click Skill install recipes, Control UI status filtering, Slack rich replies, a CLI --container flag for in-container execution, Discord automatic thread naming, a before_dispatch plugin hook, and sandbox media fixes, and states that all tutorial chapters were synced. The 2026-03-23 entry describes OpenClaw 3.22 as a major version with a plugin SDK refactor that deprecates the old extension-api, security hardening for SMB credential leakage, environment variable injection and Unicode spoofing, GPT-5.4 becoming the default, Feishu interactive cards, Telegram topic auto-naming, and agent timeouts extended to 48 hours. Two things follow. First, the tutorial is chasing a fast-moving upstream, so any chapter you read may describe a version older than the one you installed. Second, the plugin SDK refactor means older third-party extension code and older tutorial snippets may not run against current OpenClaw. If you are on the build track, check whether the chapter you are reading predates 2026-03-23 before copying plugin code.
Licence terms constrain reuse more than the code does
The repository metadata supplied here lists the primary language as JavaScript and the licence as unknown, but the README badge declares CC BY-NC-SA 4.0. Those two signals disagree, and the discrepancy matters because this is a documentation project rather than a library. CC BY-NC-SA carries a non-commercial restriction and a share-alike requirement on derivatives, which is a very different proposition from a permissive software licence. If you are a company planning to adapt these chapters into internal onboarding material, or to translate them into another language and publish the result, the non-commercial and share-alike clauses are the ones to read carefully. Nothing here is legal advice; open the LICENSE file in the repository and read the actual text rather than the badge. For individual learners following along, the licence is unlikely to be a practical constraint at all. For anyone repackaging the content, it is the first thing to resolve.
Where this tutorial is the wrong tool
Three limits are visible from the material itself. The first is language. The README, the chapter tables and the online book are Chinese-first, with English and Japanese README variants offered as badges but no indication in the supplied material that the chapter bodies exist in English. If your team reads English only, you are reading READMEs, not the tutorial. The second is that this is a tutorial, not a specification. The chapter list names topics like prompt system and message loop, but nothing in the supplied material suggests an API reference, a changelog of breaking changes with migration steps, or versioned documentation per OpenClaw release. When a plugin SDK refactor deprecates an old API, you find out from a dated changelog line, not from a deprecation guide. The third is operational. The security chapter exists and covers threat modelling, VM isolation and trust boundaries, but a tutorial chapter is not a hardened deployment. If you are putting an agent with access to your chat platforms and possibly your filesystem into a production environment, a chapter that explains MITRE ATLAS is a starting point for your own threat model, not a substitute for one. The repository also has no retrieved releases, so there is no versioned artefact to pin your reading to.
The alternative worth comparing: OpenClaw's own documentation
The obvious alternative is the upstream project's own documentation and release notes. The difference in approach is structural. Upstream documentation describes the current state of the software: what a flag does today, what a config key accepts, what changed in the latest release. hello-claw describes a path through the software: install this way, then configure this, then extend this, with a scenario library to show what the result is for. Upstream will tell you that before_dispatch is a plugin hook; hello-claw will tell you which chapter to read first if you have never opened a terminal. That makes them complements rather than substitutes, and the repository's own changelog behaviour supports reading it that way, since it re-syncs chapters against upstream releases. A second alternative is the sibling Datawhale project linked from the README, easy-vibe, which the README points to for readers who also want to learn vibe coding. That is a different subject, not a competing treatment of the same one. If you already run OpenClaw daily and want to understand its internals, the build track is the part of hello-claw that upstream documentation is least likely to give you, because it is written as a dissection exercise with alternatives considered, not as a reference.
Who should start here, and what to check first
Start with hello-claw if you are a Chinese-speaking reader who wants OpenClaw running and does not want to assemble the path yourself. The routing advice in the README is worth following literally: adoption track first, university track for a scenario you actually have, build track only when you intend to modify something. If you are a developer who wants to understand agent architecture by reading a working system's prompt and tool layers, the build track chapters on the prompt system, tool system and message loop are the specific reason to be here. Do not start here if you need English chapter bodies, a stable API reference, or a deployment you can hand to a security team unchanged. Before you invest a weekend, do three checks. Open LICENSE and confirm the CC BY-NC-SA 4.0 terms against your intended use. Pick the chapter you need and find its date in the changelog, then compare that date against the OpenClaw version you have installed, because the 2026-03-23 plugin SDK refactor is exactly the kind of change that leaves older snippets stale. And read the security chapter before you connect a chat platform, not after, since the threat model material is only useful while the trust boundaries are still yours to choose.
Editorial conclusion
Adopt hello-claw if you want a structured Chinese-language path into OpenClaw and you are willing to follow the repository's own chapter order rather than skim for commands. Skip it if you need an English-first reference, an API-level spec, or a turnkey deployment. Before committing, open the LICENSE file to confirm the CC BY-NC-SA 4.0 terms, check which OpenClaw version the chapter you need was last synced against, and read the security chapter alongside the install chapter rather than after it.
Community notes