# Get-Started-with-Web3: a bilingual curriculum that doubles as an AI-readable knowledge layer

> The repository ships 11 modules, 62 lessons and 124 indexed bilingual entries, plus llms.txt and a read-only MCP server so agents can query the same content humans read.

**beihaili/Get-Started-with-Web3** — Open-source bilingual AI-native Web3 curriculum: wallets, Bitcoin, Ethereum, DeFi, L2, DAO, smart accounts, llms.txt and MCP

- Repository: https://github.com/beihaili/Get-Started-with-Web3
- Website: https://beihaili.github.io/Get-Started-with-Web3/
- Stars: 615 · Forks: 61
- Language: JavaScript
- License: MIT
- Published: 2026-09-10 · Updated: 2026-09-10 · Language: en
- Canonical page: https://hysenlabs.com/projects/beihaili-get-started-with-web3

## The scattered-tutorial problem this curriculum tries to collapse

Beginner Web3 material lives in three places that never talk to each other: wallet setup guides, protocol documentation, and security warnings about the scams you meet on day one. Get-Started-with-Web3 puts those in one ordered path. The README states the goal directly: it turns the mess into a structured learning path that can be read by humans and queried by AI agents.

The audience table is unusually honest about who each track serves. Beginners get wallet creation, a first transaction, scam avoidance. Builders move from concepts to DApps, smart contracts, block explorers, Bitcoin RPC, DeFi, L2 and DAO tooling. Researchers get Bitcoin, Ethereum, DeFi, L2, DAO, bridge and stablecoin concepts in one place. The fourth row is the one that separates this from a typical course site: AI agents are listed as a first-class audience, with the REST API, llms.txt, JSON artifacts and MCP named as the surfaces they use.

That framing has a cost. A curriculum written for both a person skimming on a phone and a model retrieving context has to keep its structure machine-legible, which is why the repository carries an ai/ directory, a content index and a publishing step in its build. If you only want prose lessons, you are paying for infrastructure you will not use.

## How the course map and the AI index are generated

The curriculum is organized as modules with lesson counts, and the numbers are explicit in the README: 11 modules, 62 lessons in the React course map, 124 indexed bilingual lesson entries, and a 63-term glossary. The module list runs from Web3 Quick Start (7 lessons) through Bitcoin cryptography and consensus, Web3 Builder Lab (6 lessons covering ERC-20 deployment and a first DApp), DeFi Deep Dive (5), Cross-Chain and Layer 2 (6), DAO governance (5) and Ethereum with smart accounts (3).

The mechanism that keeps the site and the agent surfaces in sync is the npm script chain. package.json wires a predev hook that runs sync-content before vite starts, and a prebuild hook that runs sync-content, then ai:index, then ai:publish. So the content index and the published AI artifacts are regenerated from the same sources as the site, not hand-maintained alongside it. A postbuild step generates OG images, a sitemap and prerendered pages.

Two scripts exist purely to check that this pipeline did not drift: ai:verify runs verify-ai-entrypoints.mjs, and translation:check runs check-translation-coverage.mjs. The second one matters because the README lists languages as "Chinese first, English in progress." That is a real constraint, not a marketing line. If you need English-only material, the coverage checker is the tool that tells you what is actually translated.

## Installing it locally and running the platform

The repository is a Vite and React application, so the local path is the usual one. Clone the repository, install dependencies, and copy the environment template. The .env.example file is written in Chinese and asks you to copy it to .env.local.

```bash
git clone https://github.com/beihaili/Get-Started-with-Web3.git
cd Get-Started-with-Web3
npm install
cp .env.example .env.local
```

The AI Tutor needs a Gemini API key, which is the only required variable in the template. The file points to Google AI Studio for the key and names the variable VITE_GEMINI_API_KEY. The GitHub variables below it are marked optional with defaults, so you can leave them as they are.

```bash
# .env.local
VITE_GEMINI_API_KEY=your_gemini_api_key_here
VITE_GITHUB_USERNAME=beihaili
VITE_GITHUB_REPO=GetStartedWithWeb3
VITE_GITHUB_BRANCH=main
```

Then start the dev server. Because of the predev hook, this first runs the content sync, so the lesson data is rebuilt before Vite serves the app. You should see the sync output, then the Vite dev server URL in the terminal.

```bash
npm run dev
```

If you want to exercise the agent surfaces instead of the UI, two more scripts are available. api:dev runs serve-web3-api.mjs to bring up the REST API locally, and mcp:web3 runs web3-mcp-server.mjs to start the MCP server. Both are documented in the repository as separate entry points rather than parts of the site build.

## The REST API, llms.txt and the read-only MCP server

The agent-facing layer is the part most course repositories do not have. The README lists a public REST API documented in docs/api.md, a production base URL at get-started-with-web3.vercel.app/api/v1, and an OpenAPI 3.1 document at /api/v1/openapi.json. Alongside those sit three static artifacts: llms.txt, an AI manifest at ai/manifest.json, and a content index at ai/content-index.json.

The MCP server is explicitly described as read-only. That is a deliberate boundary rather than an oversight: an agent can search, read and cite course content, but cannot write to the curriculum through that channel. For anyone evaluating whether to wire this into an assistant, read-only is the safer default and also the limit you should expect. If your workflow needs an agent to annotate lessons or submit corrections, the MCP surface will not do it, and the README does not describe a write path.

The versioned API is the more interesting commitment. A version prefix in the URL means lesson content can change without breaking a consumer that pinned v1, which matters when the consumer is a script rather than a person. The README does not document a deprecation policy for that version, so treat the stability of v1 as asserted by the URL scheme, not by a published guarantee.

## Where this is the wrong tool

The repository draws its own boundary and it is worth taking at face value. The README states the project is a curriculum and AI-readable knowledge layer first, and that the labs it adds are educational examples, not a production dApp starter kit. If you are looking for scaffolding to ship a contract to mainnet, this is the wrong repository, and the contracts/ directory should not change that judgement.

There is a second boundary in the same section: the project is not investment advice, token promotion, exchange onboarding or a virtual-currency business service. The README adds that wallet, SIWE, x402, certificate, sponsor and paid-tool materials must be labeled as demos, metadata, drafts or future plans until the relevant production systems are live and verified. That is a disclosure policy, and it also tells you that some of the material you encounter is explicitly not production-ready.

The language split is the third limitation. The README lists Chinese first and English in progress, so an English-speaking learner may hit untranslated lessons. The translation:check script exists to measure that gap, but the README does not promise a completion date. Anyone who needs a fully English curriculum should run the checker before investing time.

## How it compares with a video course or a docs-only site

The obvious alternative is a video course platform. The difference is not production quality, it is retrievability. A video course cannot be queried by an agent, cannot be diffed in git, and cannot expose a versioned endpoint. This repository ships a content index and an OpenAPI document, which means a script can pull a lesson and cite it. If your learning happens entirely by watching, that advantage is worth nothing to you, and a video course will cover the same introductory ground with less setup.

The second alternative is protocol documentation itself, such as the Bitcoin and Ethereum references the curriculum points at. Documentation is authoritative and current; a curriculum is curated and can lag. The trade-off runs the other way too: docs assume you already know what to look for, while this project orders the material and adds the security and scam-awareness framing that protocol docs leave out.

A third comparison is a plain static site of Markdown lessons. That would be simpler to host and would need no build chain. What it would lose is the pipeline: the prebuild sequence that regenerates the AI index and publishes artifacts, and the verification scripts that catch drift. Whether that machinery earns its keep depends on whether you actually consume the agent surfaces.

## Maintenance, licence and what an upgrade costs you

The repository is not archived, and the last push was on 2026-09-07, which is recent enough that the project is being touched. Two releases are listed: interactive-learning-2026-05-18, which the release name ties to a Merkle Builder and a Gas Fee Calculator, and l2-english-ai-index-2026-05-20, described as a Layer 2 English coverage and AI-native index update. Those names suggest the English gap and the L2 module are both active workstreams rather than finished ones.

The licence is MIT, stated in the README badge and present as a LICENSE file at the repository root. MIT is permissive, so reuse and modification are broadly allowed, but the repository also carries CONTRIBUTING.md, CONTRIBUTING.en.md, a CODE_OF_CONDUCT.md and a commitlint configuration, which means contributions are expected to follow a defined process. None of this is legal advice; read LICENSE yourself if you plan to redistribute the content.

Upgrade cost is mostly the build chain. The prebuild hook runs three scripts in sequence, and a failure in ai:index or ai:publish will stop the build before Vite runs. If you fork the project, that coupling is the thing to understand first: lesson content, the AI index and the published artifacts are one pipeline, and changing the content format means touching the scripts that generate the other two. The README does not document a rollback path for a failed publish step.

## Conclusion

Adopt it if you are a beginner who wants a structured path from wallet creation to DeFi and L2, or a builder who wants course context that an agent can query without scraping HTML. Do not adopt it as a production dApp starter kit; the repository states its labs are educational examples. Before you commit, check the English coverage with npm run translation:check, read docs/api.md for the versioned endpoints, and confirm the licence terms in LICENSE.

## FAQ

### Where can I learn Web3 for free with Get-Started-with-Web3?

The project hosts a live site at beihaili.github.io/Get-Started-with-Web3 and the repository is MIT licensed, so the curriculum itself is open. It contains 11 modules and 62 lessons covering wallets, Bitcoin, Ethereum, DeFi, Layer 2, DAO and smart accounts.

### Can I learn Web3 without coding using Get-Started-with-Web3?

The Web3 Quick Start module covers wallets, a first transaction, DApp interaction, useful tools, token launch, security and CEX basics, which does not require writing code. Later modules such as Web3 Builder Lab move into ERC-20 deployment and a first DApp, so the curriculum starts non-technical and becomes technical.

### How do I get started with Web3 development using this repository?

Clone the repository, run npm install, copy .env.example to .env.local and set VITE_GEMINI_API_KEY, then run npm run dev. The predev hook syncs content first, so the lesson data is rebuilt before the dev server starts. For the API locally, run npm run api:dev.

## Sources

- [beihaili/Get-Started-with-Web3 on GitHub](https://github.com/beihaili/Get-Started-with-Web3)
- [License: MIT](https://github.com/beihaili/Get-Started-with-Web3/blob/main/LICENSE)
- [Project website](https://beihaili.github.io/Get-Started-with-Web3/)
- [README](https://github.com/beihaili/Get-Started-with-Web3/blob/main/README.md)
- [Releases](https://github.com/beihaili/Get-Started-with-Web3/releases)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/beihaili-get-started-with-web3