# AI Agent Course Map: Learn First, Then Decide Whether to Get a Certificate

> [繁體中文](./courses.md) | [简体中文](./courses.zh-Hans.md) | **English**

<!-- freshness: canonical=resources/courses.md; verified_on=2026-08-29; scope=course-availability,cost,certificate,assessment,repository-status; max_age_days=90 -->

This page helps you do one thing: **choose one course that truly fits you right now.** You do not need to collect certificates first or sign up for five courses at once. Choose one, build something you can show, and then take the next step.

To follow this project and build step by step, return to the [main learning path](../README.en.md); for unfamiliar terms, see the [glossary](glossary.en.md).

## 🧩 First, separate five terms that are easy to mix up

<div align="center" markdown="1">

| Core term | The five-year-old version | What it actually means |
|:---|:---|:---|
| **Course** | A teacher has laid out a learning path. | A set of videos, articles, exercises, or projects arranged in sequence. |
| **Certificate of Completion** | Proof that you made it through the course. | Proof that you completed the assigned content; it is not a degree and does not by itself prove that you can build a production system. |
| **Skill Badge** | A little sticker showing that you completed a task. | A digital badge issued by a platform for a short module or a specific skill. |
| **Professional Certificate** | Several courses packaged as one bigger learning bundle. | A certificate for a series of courses designed by a company or school; it is usually still not a degree or a license. |
| **Certification Exam** | You do more than take the course—you also sit another exam. | An exam through which a vendor or testing organization verifies knowledge of a specific product; it may require payment, identity verification, or periodic renewal. |

</div>

**The most important rule: a certificate proves that you completed a path; your work shows other people what you can do.**

## ⚡ Pick one path first—do not read everything at once

<div class="reader-course-picker" markdown="1">

<div align="center" markdown="1">

| What you want to do now | Start with this | Why |
|:---|:---|:---|
| You have no idea where to start | [Hugging Face AI Agents Course](https://huggingface.co/learn/agents-course) | Free; it teaches Agent, tools, and basic frameworks first, then gives you a challenge. |
| You want lots of executable code | [Microsoft AI Agents for Beginners](https://github.com/microsoft/ai-agents-for-beginners) | Every lesson includes text, video, and code; the examples lean toward Microsoft Agent Framework. |
| You want a complete Chinese-language textbook | [Datawhale Hello-Agents](https://github.com/datawhalechina/hello-agents) | Goes from principles all the way to RAG, Multi-Agent, MCP, and deployment. |
| You want to learn framework-neutral design methods first | [DeepLearning.AI Agentic AI](https://www.deeplearning.ai/courses/agentic-ai) | Builds reflection, tool use, planning, eval, and multi-agent systems from scratch. |
| You have decided to use Claude / LangGraph | [Claude Academy](https://academy.claude.com/) / [LangChain Academy](https://academy.langchain.com/courses/intro-to-langgraph) | Learn the vendors' current tools directly; remember to separate general concepts from product-specific buttons. |
| You want to organize the rules and tools given to a coding agent | [Hugging Face Context Course](https://huggingface.co/learn/context-course/unit0/introduction) | Use **context engineering**—organizing knowledge so an agent can find the right material when needed. |
| You know Agent foundations and want deeper task decomposition | [LangChain Project: Deep Agents](https://academy.langchain.com/courses/deep-agents-with-langgraph) | Use a work list, files, and **Subagents**—agents assigned part of a larger job—to complete a fuller Agent. |
| Your main goal is a certificate series | [IBM](https://www.coursera.org/professional-certificates/ibm-rag-and-agentic-ai) / [Vanderbilt](https://www.coursera.org/specializations/ai-agents) | These are longer paid course series; first confirm that the cost, language, and projects fit your needs. |

</div>

</div>

## 🎯 Selected courses and learning paths

The stars are this project's **editorial recommendation**, not a certificate ranking: ⭐⭐⭐⭐⭐ works well as a main path; ⭐⭐⭐⭐ is well worth taking but leans more toward a specific tool or purpose; ⭐⭐⭐ suits people who have already decided to use that vendor.

<small>New-course verification: 2026-10-15 UTC; scope limited to Hugging Face Context Course and LangChain Project: Deep Agents.</small>

<div class="reader-resource-catalog" role="region" tabindex="0" aria-label="Courses and learning resources">
<table align="center">
  <thead><tr><th scope="col">Group</th><th scope="col">Project / document</th><th scope="col">Rating</th><th scope="col">Use</th><th scope="col">Limits</th></tr></thead>
  <tbody>
    <tr><th scope="rowgroup" rowspan="3" align="center" valign="middle" width="80">Free foundations</th><td><div data-field="entry"><strong><a href="https://huggingface.co/learn/agents-course">Hugging Face — AI Agents Course</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> Primarily English; free</div><div data-field="outcome"><strong>Learn or build:</strong> Learn the Agent Loop and build with smolagents, LlamaIndex, LangGraph, Agentic RAG, and Eval.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> A score of 80% on the Unit 1 quiz earns a basic Certificate of Completion; the full path also includes assignments and a final challenge.</div></td></tr>
    <tr><td><div data-field="entry"><strong><a href="https://github.com/microsoft/ai-agents-for-beginners">Microsoft — AI Agents for Beginners</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> Multiple languages; free and open source</div><div data-field="outcome"><strong>Learn or build:</strong> Use text, video, and Python / .NET examples to build tools, memory, planning, RAG, Multi-Agent systems, and deployment.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> No Certificate of Completion; the current examples lean toward Microsoft Agent Framework and Foundry.</div></td></tr>
    <tr><td><div data-field="entry"><strong><a href="https://github.com/datawhalechina/hello-agents">Datawhale — Hello-Agents</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> Simplified Chinese; free and open source</div><div data-field="outcome"><strong>Learn or build:</strong> Go from Agent principles and classic patterns to RAG, memory, Multi-Agent, MCP, and complete projects.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> No Certificate of Completion; there are many chapters, so choose based on your own question—you do not need to read them all at once.</div></td></tr>
  </tbody>
  <tbody>
    <tr><th scope="rowgroup" rowspan="6" align="center" valign="middle" width="80">Build and ship</th><td><div data-field="entry"><strong><a href="https://www.deeplearning.ai/courses/agentic-ai">DeepLearning.AI — Agentic AI</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> English; videos are free to audit</div><div data-field="outcome"><strong>Learn or build:</strong> Build reflection, tool use, planning, Multi-Agent systems, error analysis, and component Eval from scratch.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> Quizzes, graded assignments, and the certificate require Pro; free auditing does not include a certificate.</div></td></tr>
    <tr><td><div data-field="entry"><strong><a href="https://academy.claude.com/">Claude Academy</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> English; free</div><div data-field="outcome"><strong>Learn or build:</strong> Learn Claude API, Claude Code, MCP, Agent Skills, and Subagents as needed.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> Passing the course quiz can earn a free completion badge; this is a Claude product path and does not replace general Agent foundations.</div></td></tr>
    <tr><td><div data-field="entry"><strong><a href="https://huggingface.co/learn/context-course/unit0/introduction">Hugging Face — Context Course</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> English; course material is free, but external API or Agent subscriptions may cost</div><div data-field="outcome"><strong>Learn or build:</strong> Turn reusable instructions into a <strong>Skill (a rule bundle)</strong> and connect tools through <strong>MCP (a tool-connection protocol)</strong>. Add <strong>Plugins (tool bundles)</strong> and <strong>Subagents (delegated agents)</strong>. Use <strong>Hooks (event-triggered actions)</strong>, then inspect a <strong>Nano Harness (a tiny execution harness)</strong>.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> The official page lists two certificate levels, but the advanced certificate requires a capstone still to be announced in a live stream; do not treat it as currently completable.</div></td></tr>
    <tr><td><div data-field="entry"><strong><a href="https://academy.langchain.com/courses/intro-to-langgraph">LangChain Academy — Introduction to LangGraph</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> English; free</div><div data-field="outcome"><strong>Learn or build:</strong> Connect steps (graph), keep task data (state), and use memory. Add <strong>human review or decisions (Human-in-the-Loop, HITL)</strong> and <strong>reusable small workflows (subgraph)</strong>, then practice deployment and long-term memory.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> Leans toward LangGraph / LangSmith; the current public course page does not clearly list certificate requirements.</div></td></tr>
    <tr><td><div data-field="entry"><strong><a href="https://academy.langchain.com/courses/deep-agents-with-langgraph">LangChain Project: Deep Agents</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> English; free course, but model/API usage may cost; 30 lessons and about 2 hours of video</div><div data-field="outcome"><strong>Learn or build:</strong> Use a <strong>Todo list (work list)</strong> to break down planning, then add files and Subagents to assemble a complete Agent.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> For readers who know Agent foundations; it leans toward LangGraph / Deep Agents, and the public page does not state a certificate threshold.</div></td></tr>
    <tr><td><div data-field="entry"><strong><a href="https://www.kaggle.com/learn-guide/5-day-agents">Google × Kaggle — 5-Day AI Agents Intensive</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> English; free self-study</div><div data-field="outcome"><strong>Learn or build:</strong> Understand Agent through models, tools, orchestration, memory, and Eval, then build a capstone.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> Originally a time-limited intensive, now used as a self-study guide; do not treat cohort participation eligibility as a permanent certificate.</div></td></tr>
  </tbody>
  <tbody>
    <tr><th scope="rowgroup" rowspan="2" align="center" valign="middle" width="80">Longer course series</th><td><div data-field="entry"><strong><a href="https://www.coursera.org/professional-certificates/ibm-rag-and-agentic-ai">IBM — RAG and Agentic AI Professional Certificate</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> English; paid subscription, financial-aid options available</div><div data-field="outcome"><strong>Learn or build:</strong> Complete RAG, Agentic AI, tools, vector databases, and hands-on projects across multiple courses.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> IBM / Coursera series certificate; it is not a degree, and cost and available assistance depend on region and account.</div></td></tr>
    <tr><td><div data-field="entry"><strong><a href="https://www.coursera.org/specializations/ai-agents">Vanderbilt — AI Agent Developer Specialization</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> English; paid subscription, with multilingual subtitles on the platform</div><div data-field="outcome"><strong>Learn or build:</strong> Use Python, tools, memory, and Agent architecture to build a set of application projects.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> Vanderbilt / Coursera Specialization certificate; some content leans toward OpenAI tools.</div></td></tr>
  </tbody>
  <tbody>
    <tr><th scope="rowgroup" rowspan="2" align="center" valign="middle" width="80">Chinese vendor tracks</th><td><div data-field="entry"><strong><a href="https://www.nvidia.cn/training/certification/generative-ai-llm-learning-path/">NVIDIA — Chinese Agentic AI learning path</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> Simplified Chinese; self-study and instructor-led courses are mostly paid</div><div data-field="outcome"><strong>Learn or build:</strong> Learn RAG Agent, Agentic AI applications, evaluation, and production deployment in sequence.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> Some courses award DLI training certificates; prices and scheduling follow the official pages, and the content leans toward the NVIDIA stack.</div></td></tr>
    <tr><td><div data-field="entry"><strong><a href="https://edu.aliyun.com/certification/cldm02">Alibaba Cloud — Build intelligent-agent applications on the Bailian platform</a></strong></div></td><td align="center" valign="middle"><span data-field="rating">⭐⭐⭐</span></td><td><div data-field="cost"><strong>Language and cost:</strong> Simplified Chinese; currently available to activate for free</div><div data-field="outcome"><strong>Learn or build:</strong> Configure intelligent agents and workflows on the Bailian platform, practice agent orchestration, and connect them to websites or DingTalk.</div></td><td><div data-field="limits"><strong>Certificate / limitations:</strong> Receive a Clouder certificate after completing the learning and exam; you must meet the identity-document conditions listed by the official site, and it is tied to the Bailian platform.</div></td></tr>
  </tbody>
</table>
</div>

For a Chinese companion to DeepLearning.AI, continue with the [open-source Datawhale guide](https://github.com/datawhalechina/agentic-ai).

## 🧪 Leave the same piece of work evidence after every course

Do not just download a PDF. Copy this small card and record checkable results for every course:

```text
The problem I solved:
Tools my Agent can use:
How I know it worked:
How it stops safely when it fails:
Runnable code or Demo link:
```

The smallest project can be an Agent that looks up information, calls one tool, and records an Eval result. When you finish, return to this project and compare it with [Stage 3 Tool Use](../stages/03-tool-use-and-hello-agent.en.md), [Stage 4 Workflow Graph](../stages/04-agent-frameworks.en.md), and [Stage 7 Production Engineering](../stages/07-multi-agent-production.en.md).

<details markdown="1">
<summary>📜 Expand: What does a certificate actually prove?</summary>

1. **A Certificate of Completion only proves that you completed the assigned steps.** It is not a degree and does not guarantee that you can independently ship an Agent.
2. **A certification exam is different from a Certificate of Completion.** It may require proctoring, identity verification, and a separate fee; do not write a free course badge as a professional license.
3. **Free does not mean poor, and paid does not guarantee a good fit.** First check whether the course includes exercises, Eval, projects, and current documentation.
4. **Put your work on your resume too.** Honestly write what you completed, what you built, and how you tested it; do not just paste a row of badges.
5. **Courses change.** Before signing up, reconfirm the cost, certificate requirements, language, and any API / cloud accounts you need.

</details>

<details markdown="1">
<summary>🔎 Expand: How is this list maintained?</summary>

- Check the course or vendor's official page first, then the official repository; third-party articles are leads only.
- Ratings assess teaching value, completeness of hands-on work, update status, and transferability—not which certificate is most likely to get you a job.
- Do not list items that have only a marketing page, an unverifiable syllabus, or a generic completion certificate presented as a license.
- Repository stars are used only to discover community attention and are not written into the main text; maintenance status comes from whether the repository is archived, its recent updates, and its current documentation.
- When costs, certificates, or cohorts change, update all three locales and their tests together.

- On 2026-10-02, W&B AI Engineering: Agents was temporarily removed from current recommendations. The old link returned 404 and the replacement could not be opened. This does not establish permanent discontinuation; reconsider after a working official course entrance is confirmed.

</details>

<small>Data verified: 2026-08-29 UTC.</small>