Easy-Vibe
Learn AI coding from zero and ship real, demoable products with vibe coding.
- Source repo
- datawhalechina/easy-vibe
- Stars
- ★ 20k
- Last updated
- 1mo ago
- Primary language
- JavaScript
- FA score
- 58/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changesClaude CodeOpenAI API · Claude API (Partial support)
- Cost
- Free, no paid service needed
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A non-technical product manager or founder who wants a demoable prototype and user feedback within a week.
- Not a fit if
- Developers looking for a plug-and-play agent framework
- Teams needing a fully offline workflow with no AI IDE
- Learners who want a single-language course rather than a 10-language site
- Source review
- 58/100 · Major gaps
What does this agent do, and when should you use it?
Easy-Vibe is a Datawhale course site for AI-native product building, built with Next.js and published as a static site at datawhalechina.github.io/easy-vibe. Content runs through three stages: Stage 1 takes beginners from spotting a real problem and validating it to building prototypes and integrating text, image and video AI features; Stage 2 covers design-to-code frontends, Supabase databases, backend APIs, Zeabur deployment and Stripe billing; Stage 3 goes deep on Claude Code — MCP, Skills, Agent Teams, long-running tasks and Spec Coding — plus cross-platform projects for WeChat Mini Programs, Android, iOS, PWA, Electron and a VS Code extension. An appendix knowledge base advertises 9 areas and 80+ interactive topics with animated, clickable explanations of RAG, Git and IDE workflows. The full tutorial exists in 10 languages, and an llms.txt file helps tools such as Claude, Cursor and Trae navigate the repo. It is a curriculum and reference, not an installable agent runtime.
The repository is mostly documentation content plus interactive components. A reader opens tutorial pages in a browser, follows the learning map into each chapter, and clicks through animated appendix pages that illustrate how RAG data flows, how Git works, or how an IDE workflow proceeds. The lessons then direct the reader to use their own AI IDE (VS Code, Cursor, Trae) to generate a project in natural language and walk it through Supabase, backend APIs, Zeabur deployment and Stripe payments; Stage 3 shows configuring MCP servers, packaging Skills, and organizing Agent Teams and long-running tasks in Claude Code. The repo itself ships npm install / npm run dev instructions for local preview and an llms.txt index so coding agents can locate relevant lessons. The output is a learning path and follow-along projects, not tasks executed by repo code.
- A non-technical product manager or founder who wants a demoable prototype and user feedback within a week.
- A student working through Stage 2 to ship a SaaS-style full-stack app with login, database, payments and an admin panel alone.
- A mid-level developer who wants structured practice with Claude Code MCP servers, Skills, Agent Teams and long-running task design.
- An indie developer comparing delivery targets — WeChat Mini Program, iOS, Android, PWA, Electron, VS Code extension — before committing.
- An instructor or training organizer who needs a multilingual reference curriculum to assign to a cohort.
How do you install or deploy this agent?
Nothing to install for reading — open the hosted site directly. To run the site locally, the README's traditional path requires Node.js and npm:
How do you use this agent?
Read online: start at the welcome page or the learning map, try the 'if you can speak, you can code' exercise, then move into Stage 1 for idea validation, prototyping and AI feature integration. Locally, either tell your AI IDE 'Please help me run this project locally' in its chat window, or start the dev server yourself and open http://localhost:3000. From there you follow the chapters inside your own AI IDE to create projects; Stage 2's SaaS example, for instance, has you connect Supabase, Zeabur and Stripe accounts and keys as described on each chapter page. Coding agents can use the repository's llms.txt to find relevant lessons.
What are this agent's strengths and limitations?
- Explicitly tiered paths — Stage 1 prototypes, Stage 2 full stack, Stage 3 Claude Code and cross-platform — so you can enter at your own level instead of reading linearly.
- Aims at real delivery: lessons include Supabase, Zeabur deployment, Stripe billing and a complete SaaS capstone, well beyond prompt-only tutorials.
- Animated, interactive appendix pages make abstract topics such as RAG, Git, diffusion models and IDE workflows easier to absorb than prose alone.
- Full 10-language content plus an llms.txt index that AI coding tools can read to answer questions about the course.
- No license in the repository metadata (GitHub shows unknown) while the README states CC BY-NC-SA 4.0, which forbids commercial use — verify before adopting.
- It is not an installable agent or CLI: no programmable interface, tests or runtime behavior to audit, so its value rests entirely on content quality.
- Exercises lean on third-party accounts and keys (Supabase, Zeabur, Stripe, Dify, Claude Code), so step-by-step reproduction is exposed to those services changing.
- News entries in the README are dated 2026, an obviously anomalous timeline, making release cadence and content freshness hard to judge from the source.
How does this agent compare with similar options?
The README's Stage 1 appendix includes a comparison of seven AI coding tools and hands-on articles that use products such as Claude Code, Cursor, Trae and Dify as teaching subjects. Easy-Vibe is a course, not a substitute for those tools, so it has no direct competitor named in the source.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Easy-Vibe This agent | 58 · Major gaps | Web appFree | ★ 20k | 1mo ago | JavaScript | Claude Code |
| Vibe Vibe – AI Coding for Everyone | 22 · Major gaps | Web appFree | ★ 6.2k | 5mo ago | Dockerfile | — |
| Create Context Graph | 55 · Major gaps | CLIFree + model costs | ★ 738 | 10d ago | Python | ChatGPT · Claude Code · Claude.ai · OpenAI API · Claude API |
| AI Agents from Zero: Practical Bootcamp to Enterprise Deployment | 34 · Major gaps | CLIFree + model costs | ★ 5.4k | 5d ago | Python | OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
This is a VitePress tutorial site rather than an executing agent, so the privilege surface is naturally small: deploy.yml grants only contents:read, pages:write, id-token:write and release-books.yml grants contents:write for publishing books, which is proportionate least privilege (2). However there is no user confirmation mechanism (1), no explanation of data flows when tutorials call third-party AI/payment/database services (1), and no sensitive-data guidance (1). Dependencies use loose ^ ranges including a vitepress 2.0.0-alpha prerelease and an obscure claude ^0.1.1 package, with no lockfile audit or vulnerability notes, so dependency_security is 1. External effects are limited to build and publish and are controllable (2). Rollback is only implicit via Pages/Release artifacts with no documented recovery path (1). Source attribution is present through extensive external links and contribution templates (2).
README table of contents, learning paths and News entries are mutually consistent, and package.json scripts line up with workflow invocations (npm run build, book:all), so self_consistency is 2. Dependencies are public npm packages and the Node >=18 engine matches the workflow's Node 20, giving dependency_availability 2. There is no learner-facing failure messaging or troubleshooting mechanism beyond a single appendix link, so failure_messages is 1.
README clearly segments audiences (complete beginners, PMs/founders, students, junior and senior developers) with multiple learning paths, so audience_and_scenarios is 3. Capability boundaries are described by stage (Stage 1-3 plus appendix) but the docs do not state that tutorial outputs are not guaranteed for quality or safety, so capability_boundaries is 2. There is no agent trigger mechanism, so trigger_precision is 1. Environment fit covers local VitePress, GitHub Pages and multilingual book builds, giving environment_fit 2.
README table of contents, learning map and multilingual entry points are well organized, so information_architecture is 3. Run Locally and npm scripts give install hints, but full prerequisites and common install issues are missing, so install_notes is 2. Naming is stable (easy-vibe, stage-1/2/3), giving naming_stability 2. Examples and project cases are abundant, so examples_and_faq is 3. There is essentially no dedicated known-limitations section, so known_limitations is 1. The license is stated as CC-BY-NC-SA-4.0 in the README badge and package.json, but the task metadata says unknown, an inconsistency, so license is 2. News acts as a change log but there is no formal CHANGELOG or semantic versioning, so versioning_changelog is 2. Maintenance responsibility is implied by the datawhalechina org and contribution guide, but publisher identity is unverified, so maintenance_responsibility is 2.
Tutorial outputs are demoable prototypes and full-stack apps, so output_usability is 2. Compared with generic AI coding material, the staged roadmap and interactive demos add some marginal value (2). Cost is a free tutorial plus local build, but dependencies are heavy (Puppeteer, LaTeX toolchain), so cost_benefit is 2.
Many README claims (tens of thousands of learners, 10 languages, 80+ interactive topics) are marketing-style with no traceable source, so claim_traceability is 1. Cross-source corroboration is limited to News and table-of-contents agreement, giving cross_source_corroboration 1. Facts and inferences are largely not separated, with tutorial examples mixed with real outcomes, so fact_inference_separation is 1.
- This is a tutorial site, not an executable agent; FARS scores reflect docs and build config, not runtime safety.
- package.json uses loose ranges and includes a vitepress 2.0.0-alpha prerelease and an obscure claude ^0.1.1 package; audit the lockfile and dependency provenance.
- Tutorials involve Stripe, Supabase and third-party AI services; learners must assess keys and data flows themselves, as the repo gives no guidance.
- License is stated as CC-BY-NC-SA-4.0 in README/package.json but metadata says unknown; verify before use.
- Publisher identity is unverified; treat maintenance commitments and update paths with caution.