rudnfaiActual Inflearn course review
“I was able to understand the core principles of how agents work by implementing them myself.”핵심 원리 이해
TypeScript · Ollama 로컬 · 100% 무료
An AI agent that uses tools,
An AI agent that uses tools,
built by hand
This course is a TypeScript course where you first implement an AI agent loop by hand and then rebuild it with the Vercel AI SDK. You’ll write the agent loop once without a framework to see how it works firsthand, and then complete it again with the Vercel AI SDK. A real agent that runs on your laptop—without the cloud or paid APIs. In the final bonus section, you’ll swap out only the agent’s model for Kimi K3 and see how far you can go beyond what a local model could do.
Find the meeting schedule in the workspace and summarize it in one line.
[tool] search_files("meeting") → notes.md
[tool] read_file("notes.md")
The next meeting is on Friday at 3 PM.
calculate({"expression":"17*23"}) → 391read_file("notes.md")write_file("todo.md") → confirmation ylist_files({})search_files("회의") → notes.mdremember("likes tea")recall("like") → memory recallfetch_url("https://example.com")current_time({})stopWhen: stepCountIs(8)temperature: 0steps 2 steps · Multistep complete
calculate({"expression":"17*23"}) → 391read_file("notes.md")write_file("todo.md") → confirmation ylist_files({})search_files("회의") → notes.mdremember("likes tea")recall("like") → memory recallfetch_url("https://example.com")current_time({})stopWhen: stepCountIs(8)temperature: 0steps 2 steps · Multistep complete
Now, let me show you before I tell you.
Preview video
Instead of a hundred words of explanation, one minute is all you need.
An AI agent is a program that understands a user’s request, selects tools on its own, performs tasks in multiple steps, and then verifies the results. First, see how the finished product actually works. In response to a single-line command, the agent chooses and uses tools, then decides what to do next on its own. The text-based flow below breaks it down step by step afterward.
A complete local work agent demo · Since it uses a local model, the wording may vary slightly with each run.
Breaking down that agent we just saw, step by step
Once completed, it does things like this.
If you expand the conversation to include the tool calls
Once you turn it on, it can read and write files, perform calculations, fetch content from the web, and remember the user. Before carrying out risky operations, it asks a person for confirmation. Below is the actual conversation flow of the local work agent completed at the end of the lecture.
My name is Hong, and I like tea. Remember that.
[tool] remember("The user's name is Hong, and they like tea")
Hong, I’ll remember that you like tea.
Select only the unfinished tasks from todo.md and organize them in todo-open.md.
[tool] read_file("todo.md")
I’ll write 38 characters to 'todo-open.md'. Shall I proceed? (y/N) y
[tool] write_file("todo-open.md")
I organized the 2 unfinished tasks into todo-open.md.
What was my name again?
I'm Hong.
Tool use, multistep reasoning, memory, and human confirmation. It all runs on code written directly by you.
But how is this possible?
What This Course Promises
It does not stop at “knowing how to use it.”
There’s nothing special about it. The order is simply different—we first implement the principles by hand, then build the framework on top of them. Usually, people do it the other way around.
Before taking the course
They know how to use a framework. But they don’t know what’s happening under the hood.
→
After taking the course
Because we wrote the loop ourselves, we know what the framework is doing under the hood. Even when we encounter a new tool, we adapt quickly.
01
Once by hand, then with the SDK
You’ll first implement the function-calling loop directly in raw code to see how it works, then rebuild the same thing with the Vercel AI SDK. You’ll understand what the framework does under the hood.
02
100% Local, 100% Free
Run the model on your own computer with Ollama. No API key, payment, or GPU is required. Experiment as much as you like at zero cost. Kimi K3 in the bonus section also supports credits, so you can follow along without making any payments.
03
We’ll cover robustness too.
We don’t stop at a demo. We build agents that truly hold up in practice—with error recovery, input validation, infinite-loop prevention, and human confirmation for risky actions.
You can start with that sequence right now.
Once you write one by hand, you’ll see frameworks in a completely different light.
지금 시작하기And one more thing: we added it through a collaboration.
Moonshot AI Collaboration · Bonus Section
How far can we go by plugging K3 into the same agent?
In the agent we built through Section 7, we simply swap the model for Kimi K3. We leave the code almost entirely unchanged. We directly compare how far it opens up tasks that a small local model couldn’t handle—and what remains the same. Created together with Moonshot AI, with Kimi credits totaling 1,500 USD available on a first-come, first-served basis until August 21.
MOONSHOT AI · Collaboration Section
K3
Moonshot AI’s latest flagship model. Start by writing this name in a single line of code.
OFFICIAL K3 MODEL SPEC
2.8TTotal parametersThe world's first open 3T-class model
1,048,576Maximum context tokensLarge repositories and long documents in a single request
90.4BrowseComp scoreAt 1M context, without context compression
Native VisionImage understandingRead error screens and visual materials directly
BrowseComp 90.4 is Kimi’s officially published evaluation score. It was measured without context compression using a 1M-token context and the max reasoning intensity setting.
Kimi K3 공식 성능 근거 보기 →Kimi credits worth a total of 1,500 USD · First come, first served
Until August 21 · First come, first served · Ends when supplies run out · The application link is in the lecture
Kimi CLI is this kind of tool.
1 million tokens
Whole, without cutting it up
The context window is 1,048,576 tokens. You can insert a 3,600-line code file as-is without splitting it and ask, "Where is this value determined?"
Images · Video
The error screen as-is
It does not require OCR. If you capture a broken screen and send it as is, the model reads it and pinpoints the bug. It can be built into a tool and connected to the agent loop as well.
Agent Swarm
Parallel sub-agents
The orchestrator breaks the work down, distributes it to sub-agents, and gathers the results again. Up to 300 can run simultaneously (Beta, for members).
This is a scene where an image was inserted as-is to have it find bugs. We try it ourselves in Section 8.
It is a 2.8-trillion-parameter MoE model. We begin by reading the official blog together and distinguishing what has been empirically measured from what is merely marketing.
That sequence, building up step by step over 8 stages.
An Agent That Grows Through 8 Stages
At the end of each section, the agent grows by one stage.
Rather than viewing it as a table of contents, it’s better to look at what gets added at each stage. I’ve noted what your agent will newly be able to do after completing each section.
1
Section 1
What Is an Agent, and What Will We Build?
→ Understand the agent loop (perception, decision-making, action, and observation), and set up a local development environment.
- The True Nature of AI Agents and Agent Loops
- The Big Picture of the Concepts You’ll Learn in This Course
- Preparing the development environment and setting up Ollama and a tool-calling model
2
Section 2
Hand-Built Agent
→ Build a function-calling loop from scratch without a framework to gain a hands-on understanding of how agents work.
- The meaning of the messages array and function calls
- Defining tools, executing them, and checking the results
- Write the agent loop yourself and add more tools
- Observing Multi-Step Reasoning, Infinite Loops, and Stopping Conditions
3
Section 3
Rebuilding with the Vercel AI SDK
→ Cleanly restructure the hand-built loop using an industry-standard framework
- Introduction to the Vercel AI SDK and Ollama Integration
- generateText call, tool definitions, and Zod schemas
- Step tracking, response streaming, and multi-step patterns
4
Section 4
Local Work Agent — Tool Integration
→ Turn it into an agent that can handle real tasks by adding file, web, and calculation tools
- Tool schema design principles and a safe workspace
- File reading, writing, listing, and search tools
- Fetching Web Pages, Context-Saving Patterns
- Practical Scenarios for Chaining Multiple Tools
5
Section 5
Agent Memory
→ Build an agent that remembers users through short-term (conversation) and long-term (file) memory.
- Why Memory Is Needed and the Limitations of the Context Window
- Managing Old Context and Designing Long-Term Memory
- Demonstration of tools for storing and recalling facts, and an agent that remembers
6
Section 6
Robustness and Control
→ Build a robust agent that doesn’t break down on errors and involves a human for dangerous tasks.
- Why Local Models Go Off Track and Loop Control
- Validate inputs, return errors as results, and recover
- Behavioral rules and human confirmation patterns through system prompts
- Debugging and Handling Common Failure Patterns
7
Section 7
Capstone — A Complete Local Task Agent
→ Combine all the pieces so far to build a complete agent that runs in the terminal.
- Capstone Design, Thin REPL Harness, Loop Assembly
- Streaming polish, multi-step end-to-end demonstration
- What We Didn’t Cover and Next Steps, Wrap-Up
8
Collaboration Section · Moonshot AI
Swap the same agent to Kimi K3.
→ Evaluate what becomes possible and what remains unchanged when swapping models, factoring in costs as well.
- Read the official K3 blog together to distinguish its features from its limitations
- Run the agent built through Chapter 7 as-is with Kimi.
- Execute work workflows, coding workflows with Kimi CLI
- Throw in an image of the error screen as-is to find the bug
- Insert long code and documents in their entirety without cutting them.
- Differences from Claude · Codex, Token Cost Calculation
- Parallel execution of sub-agents with Agent Swarm
- Pricing, Adoption Decisions, and When to Use It
This is a bonus section created together with Moonshot AI, the company behind Kimi. We’re offering a total of 1,500 USD worth of Kimi credits on a first-come, first-served basis until August 21. There is no separate per-person limit, so the earlier you apply, the more generously you can receive. The offer ends when the deadline passes or the credits run out. The application link is included in the course.
If you’ve watched this far, it’s time to start with Step 1.
You can get started right away by installing just Ollama. The cost is 0 won.
수강 시작하기So what remains is…
What This Course Fills in
One agent, done properly from start to finish
We’ll go from the underlying principles all the way to completion. By the end of the course, you’ll be able to build things like these yourself.
You’ll build it yourself in this course.
- Implement the agent loop by hand yourself
- Using the Vercel AI SDK and How It Works Internally
- Function calling (tool use) and multi-step reasoning
- Integrating file, web, and calculation tools
- Short-term and long-term memory
- Error recovery · guardrails · human confirmation
- Swapping in the Kimi K3 Model (Bonus)
Then who is it suitable for?
Before you start
This is suitable for those who fit the following profile, and this is all you need to prepare.
◆Recommended for these people
- Those who are trying to properly build an AI agent for the first time
- Those who have found it frustrating to use frameworks as black boxes
- Those who want to be able to use it and understand how it works"
- Those who want to experiment freely on their local machine without worrying about costs
◆Requirements
- TypeScript / JavaScript Basic Syntax
- Experience using Node.js and npm
- macOS environment (using zsh)
- Install Ollama — CPU is sufficient without a GPU
This is a good fit for you if you’re someone who...
Instructor Introduction
A developer who likes building things from the ground up, starting with the underlying principles
SILICON VALLEY · AI ENGINEER
Silicon Valley Developer Waddy
I’m Waddy, a backend server developer with 13 years of experience. After developing at various companies in Korea, I recently moved to Silicon Valley, where I’m now working on AI-related development.
I connected with a developer friend at Toss and ended up joining this way. I’m very excited about being able to provide a wide range of information online. I’ll do my best to share the knowledge I have in the easiest and most accessible way possible.
"Anyone can use a framework. Once you build one yourself, even just once, you’ll see those tools in a completely different way. That perspective is ultimately what makes the difference in skill."
AI agents
tool use · function calling
local LLMs
Silicon Valley practitioners
Before getting started, things you may be curious about
Frequently Asked Questions
Things You May Be Wondering About First
Can beginners who have just started coding follow along as well?
You’ll need a basic understanding of TypeScript or JavaScript syntax and experience using npm. Building on that, the course covers the concepts needed for agents—LLM calls, function calls, and tool use—from the ground up.
Do I need a GPU or a paid API?
No. We run a small model on your computer with Ollama. CPU is sufficient without a GPU, and you don’t need an API key or payment. The Kimi K3 used in the bonus section works the same way. Since this section was created with Moonshot AI, Kimi credits totaling 1,500 USD are available on a first-come, first-served basis until August 21. There is no separate per-person limit, so the earlier you apply, the more generously you may receive. The offer ends when the deadline passes or the credits run out. The application link is inside the course. You will be able to see it after enrolling in the course.
What is Kimi K3, which is covered in the bonus section?
Only the model of the agent built in Sections 1–7 is swapped out for Kimi K3. We leave the code almost entirely unchanged and see what differs. K3 supports a 1-million-token context and image and video inputs, opening up possibilities that weren’t available with a small local model. For example, you can feed it an entire long code file without splitting it up, or give it a screenshot of an error and have it find the bug. We’ll read the official blog together and examine which parts are factual and where the marketing begins. This section was created in collaboration with Moonshot AI.
Don't you use frameworks like LangChain?
We use the industry-standard Vercel AI SDK. However, instead of using it right away, we first build the core loop by hand, then move it over to the SDK and compare them side by side. This way, you’ll understand even what the framework is doing under the hood.
Is it possible on Windows as well?
The course is conducted using macOS (zsh) as the baseline. The code and concepts themselves are the same, but installation and execution commands will be demonstrated for macOS.
What will I take away from the course?
You’ll come away with your own local work agent that runs in the terminal—and, above all, a solid understanding of “how agents work.” It’s a foundation that will help you adapt quickly even when you encounter other frameworks.
Is there any reason to take this course even if I’ve already used the Vercel AI SDK?
Knowing how to use an SDK and understanding what it does under the hood are two different things. This course first has you build the function-calling loop by hand, then reconstruct it with the SDK, so even if you already use the SDK, you’ll gain a fresh perspective on what’s happening inside it.
Student’s record
A student who implemented it themselves gave this review:
5.0Excerpted review · Other AI agent courses
This review is excerpted from an actual Inflearn course review for an AI agent course. It is displayed separately to avoid being mistaken for a review of the course on this page.
AI Agents—Now Build One Yourself
Open the black box and complete an agent that works with your own hands.
수강 시작하기TypeScript · Ollama Local
Even after the course ends, it’s not the end.
Growing Together
So it doesn’t end with listening alone
Learning continues beyond the course. We run a KakaoTalk open chat where developers gather to freely share career concerns, real-world work situations, and technical topics. If you get stuck while taking the course, feel free to ask questions.
KakaoTalk open chat where developers gather
From career topics and real-world work situations to technical concerns—a space where you can casually ask questions and find answers together.
Career StoriesIndustry Situation SharingTechnical DiscussionsFree for Everyone
오픈채팅 참여하기인스타그램 보러가기
open.kakao.com/o/ggBVZsai
I was interested in AI agents, but until now, I had only gone as far as following examples using LangChain or various frameworks. Because of that, I only had a vague understanding of how agents actually operate internally, why Tool Calling is necessary, and how the Agent Loop functions. The best part about this lecture was that, rather than relying on complex frameworks from the start, I was able to understand the core principles of how an agent works by implementing them myself. I had heard the term "agent" many times but was curious about exactly how they are implemented; I am very satisfied that I was able to build one firsthand. Before taking the course, agents felt like a somewhat abstract concept, but now I have a clear mental framework of how to design the structure and what needs to be controlled when creating an agent. I highly recommend this course to anyone who wants to build a single agent from scratch, from start to finish.
Thank you for such informative content. It was even more helpful because you explained the core aspects in an easy-to-understand way, rather than just covering simple topics like how to use Claude or Codex.
This curriculum first implements the tool-calling loop at the core of an agent directly in vanilla TypeScript to understand how it works, then refactors it using the Vercel AI SDK. It helps you understand what the framework handles internally and covers realistic challenges you encounter in production, such as preventing infinite loops, automatically recovering from errors, and requiring human approval for risky operations. It seems like a useful course for developers who want to control agents reliably in real-world applications.
Until now, I intentionally avoided diving deep into AI-related technologies. I prioritized other areas of learning because I believed that having a high-level understanding of my own technical stack was more important in the AI era. However, as AI evolves faster than expected, I began to feel that even simple side projects felt incomplete without some form of AI integration. Through this lecture, I easily gained insights into local LLMs and how to utilize them, enabling me to apply these technologies without worrying about costs. Up to Chapter 3, there were no difficult parts for a standard backend developer. However, I think the true charm of this course lies in learning the instructor's development philosophy and habits—things that aren't easily observed elsewhere. I'm learning a lot!
I was able to learn the principles of AI Agents by building one myself. Previously, I only used AI for asking and answering questions in a chat format, but by implementing how an AI Agent works internally from scratch, I was able to understand AI systems. It was exactly the lecture I needed. Thank you!
![[AI Cheat Sheet] The secret to finishing work instantly, Agentic AICourse Thumbnail](https://cdn.inflearn.com/public/files/courses/340717/cover/ai/1/3b5cb844-25b5-4576-8224-d293d0989376.png?w=420)
![[Inflearn Award Bestseller] How to Become an AI Automation Expert Without Coding, The Complete n8n GuideCourse Thumbnail](https://cdn.inflearn.com/public/files/courses/337053/cover/01jrkjt8emye2z7j7846br3vkk?w=420)
![[VOD] A story of a non-developer planner building AX in the fieldCourse Thumbnail](https://cdn.inflearn.com/public/files/courses/343357/cover/01ky6xa21r5qw8nveevz7317kj?w=420)
![[Integrated] Introduction to AI Video Production + Practical ApplicationCourse Thumbnail](https://cdn.inflearn.com/public/files/courses/343029/cover/01kwtvq9840gqg1t7b6krs3bkk?w=420)
![[Integrated] Introduction to AI Video + Shorts ProductionCourse Thumbnail](https://cdn.inflearn.com/public/files/courses/343177/cover/01kx2mmgmppa8h7jdb8kx9bxye?w=420)
![[CLI, MCP] Creating a Unity Idle Game with AICourse Thumbnail](https://cdn.inflearn.com/public/files/courses/343716/cover/01kzre6epavjnb1gk13svtewvq?w=420)