# Hello Claw: the Datawhale tutorial for adopting and building an OpenClaw assistant > Hello Claw is a Chinese-language VitePress tutorial that walks readers through installing OpenClaw, configuring it, and writing their own Skills. It is a book, not a tool, and that distinction decides whether it is useful to you. **datawhalechina/hello-claw** — 哈喽！龙虾 🙋‍♀️ Adopt from scratch and build your first claw 🦞 来领养你的第一只龙虾！ - Repository: https://github.com/datawhalechina/hello-claw - Website: https://datawhalechina.github.io/hello-claw - Stars: 2,211 · Forks: 229 - Language: JavaScript - License: not declared - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/datawhalechina-hello-claw ## What Hello Claw actually is, and who it is written for The name invites confusion. Hello Claw is not a claw machine, and it is not an application you download and double-click. It is a tutorial repository from Datawhale China aimed at OpenClaw, described in the README as a command-line AI assistant system. The repository holds documentation, an asset folder, a scripts folder, and a VitePress configuration; the top-level entries are README.md, README_EN.md, README_JA.md, package.json, package-lock.json, .github/, asset/, docs/ and scripts/. The README splits the curriculum into three parts. The first, called adopting the lobster, covers installation in chapters 1 to 3, core configuration in chapters 4 to 6, extended operations in chapters 7 to 9, and security plus clients in chapters 10 and 11, with appendices A through G. The second part, the lobster university, is a menu of scenario walkthroughs built around Skills selection: email assistance, local health management, morning briefings, calendar operations, Vibe Coding, CI/CD, documentation generation, content studio, revenue operations, meeting scheduling, multi-agent collaboration, a knowledge base, a one-person company, security checklists, paper pushing, smart home control, financial analysis and training support. The third part, building the lobster, is eleven chapters that take OpenClaw's source apart, survey alternatives, and then rebuild Skills, channels and a full customisation. The README names four audiences: people with no programming background who want an always-available assistant, users who want to drive the assistant remotely through QQ, Feishu or Telegram, technically curious readers interested in the skill system, and developers who want to understand the agent architecture well enough to build their own version. That last group is the one the repository is really shaped around, because the majority of the page count is implementation, not usage. ## How the tutorial is organised, and how it is built The documentation is a VitePress site. The package.json declares vitepress as a dev dependency at ^2.0.0-alpha.15, alongside tailwindcss and @tailwindcss/vite at ^4.2.1, markdown-it-mathjax3 at ^4.3.2, and @types/node at ^25.5.0. The only runtime dependency is vitepress-plugin-image-viewer at ^1.1.6. There is no application code, no server, no CLI of its own. Everything the reader learns is about OpenClaw, and everything the reader runs is either OpenClaw or the local documentation server. That structure has a consequence worth stating plainly. A VitePress site with an alpha-version core dependency is a moving target. VitePress 2.0.0-alpha.15 is a pre-release, and pinning to an alpha means a future npm install can resolve to a different alpha with different behaviour. The repository does ship a package-lock.json, which is the mitigation: installing from the lockfile reproduces the dependency tree the authors used. If you clone and run npm install without the lockfile discipline, you are testing a site the authors never built. The content itself is organised as a linear curriculum rather than a reference. Chapters are numbered and sequenced, the README recommends different entry points for different readers, and the university section is explicitly described as menu-style, meaning you pick five to ten Skills and build a workflow rather than reading front to back. The build-the-lobster part reverses the direction: it starts from OpenClaw's internals and works outward to customisation. So the three modules are not three views of the same material. They are three different reading modes, and the README says so. ## Running the site locally and starting with the AutoClaw chapter Because the repository is a documentation site, the first real use is building and serving it. Clone the repository, then install dependencies and start the development server. The package.json defines dev, build and preview scripts, plus docs:dev, docs:build and docs:preview aliases that point at the docs directory. ```bash npm install npm run dev ``` The dev script runs vitepress dev docs, so the server reads from the docs folder. VitePress prints a local URL in the terminal; open it and you get the Chinese documentation site, which is the same content published at the homepage URL in the README. The docs:dev alias does the same thing. For a static build you would run npm run build, which is vitepress build docs, and npm run preview to serve the result. Once the site is open, the README's own learning advice is the sensible starting point. Readers with no programming background are told to begin with the first part, adopting the lobster, and to get installation and basic automation working before anything else. That first part opens with chapter 1, AutoClaw one-click installation, described as downloading the AutoClaw desktop client for a five-minute, zero-threshold experience. Chapter 2 is the manual route: terminal introduction, Node.js installation, npm install, and the onboard configuration wizard. Chapter 3 covers the initial configuration wizard, including the CLI wizard, macOS guidance, a custom provider, and reconfiguration. The practical instruction this implies: if you have never used a terminal, read chapter 1 and use AutoClaw. If you have, read chapter 2 and install OpenClaw manually, because the manual path is what the later chapters assume. The README does not present AutoClaw as a substitute for the manual install; it presents it as the entry ramp. ## Where the tutorial stops being enough The most obvious limitation is language. The repository carries README files in Simplified Chinese, English and Japanese, and the README badges link to README_EN.md and README_JA.md. But the curriculum itself lives under docs/cn/, and every chapter link in the table of contents, including all of the university scenarios, points into that Chinese tree. A reader who needs English prose gets a translated README and then hits Chinese chapter content. That is not a small gap. It is the difference between a tutorial you can follow and a repository you can only skim. The second limitation is that there are no releases. The recent releases list is empty, which means there is no versioned snapshot of the curriculum, no changelog artefact, and no way to pin the tutorial to a specific OpenClaw version. This matters because the README's news section is dense with OpenClaw version movement: an entry dated 2026-03-23 describes OpenClaw 3.22 as a major release with a plugin SDK refactor that deprecates the old extension-api, security hardening, GPT-5.4 becoming the default, and agent timeouts extended to 48 hours. A tutorial that tracks a project moving that fast will lag somewhere, and without releases you cannot tell which chapter was written against which OpenClaw. The third limitation is licence ambiguity. The README badge displays CC BY NC SA 4.0 and links to a LICENSE file, but the licence field in the repository metadata is unknown. The badge is a claim; the metadata is what tooling reads. CC BY-NC-SA is a non-commercial licence with a share-alike condition, which is a very different proposition from a permissive software licence, and anyone planning to reuse the chapters in paid training material needs to resolve that discrepancy rather than assume. None of this makes the tutorial wrong. It makes it a book with a publication date, not a dependency you can rely on. ## Hello Claw against reading OpenClaw's own documentation The honest alternative is OpenClaw's own documentation and source. Hello Claw's build-the-lobster module is, by its own description, an attempt to take OpenClaw's source apart: prompt system, tool system, message loop, multi-channel access, then alternatives, then Skills, channels and full customisation. If you can read the source, the source is authoritative and always current, and the tutorial's eleven chapters are a guided summary of it. The difference in approach is real, though. OpenClaw's documentation tells you what the system does. Hello Claw tells you what to do with it, in sequence, with a named entry point for each kind of reader and a set of scenario walkthroughs that map Skills to concrete jobs. The university section is the part with no direct equivalent in a project's own docs: a menu of email assistance, meeting automation, CI/CD, knowledge-base sharing and similar workflows, each presented as something you assemble from Skills rather than something the maintainers shipped. That is editorial work, and it is the reason to read a tutorial instead of a manual. A second alternative, for readers who are only interested in the usage half, is to skip the repository entirely and read the published site at the homepage URL, which is the same VitePress output without a local Node.js toolchain. That is not a competing project; it is the same project with the build step removed, and for a non-developer audience it is probably the correct default. The repository matters when you want to read the Markdown directly, search it, or contribute. ## Maintenance, upgrade cost and licence The repository is not archived, and the last push was on 2026-08-19. The README's news entries stop at 2026-03-25, which describes a round of scenario expansion and a rewrite aimed at beginners, adding eleven hands-on cases across personal productivity, programming, content creation, business sales and multi-agent collaboration. The same entry notes OpenClaw v2026.3.24 changes, including Gateway OpenAI-compatible endpoints at /v1/models and /v1/embeddings, a Microsoft Teams SDK integration, one-click Skill installation 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 security fixes, with the note that all tutorial chapters were synchronised. That is the upgrade cost in one sentence: every OpenClaw release with breaking changes obliges the authors to revisit chapters, and the 3.22 entry shows what breaking means here, with the old extension-api deprecated in favour of a refactored plugin SDK. If you follow the build-the-lobster chapters, expect to reconcile them against your installed OpenClaw version yourself. There is no release tag to anchor that reconciliation. On licensing, the README badge declares CC BY NC SA 4.0 and links to a LICENSE file in the repository, while the repository metadata reports the licence as unknown. The non-commercial and share-alike terms are the practical constraint: internal training use and personal study sit comfortably inside that, redistributing adapted chapters inside a commercial product does not, and the badge-versus-metadata mismatch is something to check against the LICENSE file itself before you rely on either. ## Conclusion Adopt Hello Claw if you want a structured Chinese-language path into OpenClaw and are willing to follow a book rather than run a tool: start with the AutoClaw chapter if you have no terminal experience, or the manual install chapter if you do. Skip it if you need English documentation, a packaged installer, or a maintained release artefact, because the repository contains a VitePress site and no release has been published. Verify two things before committing: the licence file the README badge points to, and whether the OpenClaw version the chapters describe still matches the OpenClaw you install. The repository was last pushed on 2026-08-19 and is not archived. ## FAQ ### What is Hello Claw used for? It is a tutorial for OpenClaw, a command-line AI assistant system. The README divides it into adopting the assistant (installation, configuration, operations, security), scenario walkthroughs built around Skills, and a section that takes OpenClaw's source apart so you can build your own version. ### How do I install AutoClaw according to Hello Claw? The README's chapter 1 is titled AutoClaw one-click installation and describes downloading the AutoClaw desktop client for a five-minute, zero-threshold experience. The repository itself gives no AutoClaw download command or package name; the instructions are inside that chapter on the published site. ### What is Claw AI used for in this tutorial? The README describes OpenClaw as a command-line AI assistant that can be driven remotely through QQ, Feishu or Telegram, extended with Skills, and scheduled with cron-style tasks. The tutorial's university section maps those capabilities to concrete workflows such as email assistance, meeting scheduling and CI/CD. ## Sources - [datawhalechina/hello-claw on GitHub](https://github.com/datawhalechina/hello-claw) - [Issues](https://github.com/datawhalechina/hello-claw/issues) - [Project website](https://datawhalechina.github.io/hello-claw) - [README](https://github.com/datawhalechina/hello-claw/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/datawhalechina-hello-claw