7 AI agent certificates, in the order to do them

Seven days, seven credentials

1. Day 1 · Anthropic, Building effective human-agent teams. 45 min. Completion badge. No coding.
2. Day 2 · LangChain, LangChain Essentials Python. 18 lessons, 1 hour of video. Certificate of completion. Python needed.
3. Day 3 · Anthropic, Claude Platform 101. 1.5 hr. Completion badge. Coding.
4. Day 4 · Anthropic, Model Context Protocol: Advanced topics. 1.5 hr. Completion badge. Coding.
5. Day 5 · LangChain, LangSmith Fleet, Build No-Code Agents. 9 lessons, 1.5 hours of video. Certificate of completion. Explicitly no-code.
6. Day 6 · Microsoft, Develop an agent with integrated tools. 45 minute timed lab, Intermediate. Microsoft credential. Coding.
7. Day 7 · MongoDB, Building AI Agents with MongoDB. 75 minutes plus a 10-question skill check. Credly badge.

What "free" means on each of these, honestly

There are four different systems here and they each mean something slightly different by free.

• Anthropic, days 1, 3 and 4. No course page on Claude Academy prints a price. The proof is on the Claude Academy FAQ, which states it outright: "Academy certificates of completion (badges) are free and are earned by passing a course's quizzes on academy.claude.com. The Claude certification exams are a separate paid, proctored program." Every one of these three carries a "Completion badge" chip in its header, and on several courses the badge is the last item in the curriculum list, right under the course quiz. So the thing you earn here is a completion badge, not a certificate, and Anthropic's paid proctored certification exams are a separate product you cannot simply sit.
• LangChain, days 2 and 5. The price field says "Free" in plain words on both. The certificate is real, and the proof is the certificate LangChain Academy itself issues you: each one is its own page naming the course, with an issue date and a certificate ID. Neither course page mentions it though, which is why most people never collect theirs. Day 2 explains exactly where it hides, and the same trick works for day 5.
• Microsoft, day 6. The honest answer is that Microsoft publishes no price either way. Not on the credential page, not on the Applied Skills hub, not in the assessment FAQ. So this is the one day on the list where nobody, me included, can tell you Microsoft says it is free, and you should not repeat anyone who claims they can. Check the price on the credential page yourself before you commit an evening to it.
• MongoDB, day 7. The clearest of the lot, and part of why it is the finale. MongoDB states it plainly on its skill badges page: "MongoDB Skill Badges are free, focused credentials designed to help you quickly learn and validate specific MongoDB skills." The badge's own Credly page prints a Cost field reading "Free" right next to the exact requirements. Worth knowing: the course page itself will not load until you sign in, so the Credly badge is the page to read first.

Day 1: Anthropic, Building effective human-agent teams

Open the course · 5 lessons · 45 min · free, completion badge, no technical background needed

What it is. The shortest thing on the list and the right place to start, because it covers the part everyone skips: how a human and an agent actually divide work without the human becoming a full-time babysitter.

What it actually teaches. Clear roles, writing a north star the agent can be held to, releasing autonomy gradually instead of all at once, and deciding what information the agent is allowed to reach. It is aimed at "team leads and innovators whose organizations already use AI in single-player mode", which is most people reading this.

What you walk away with. A completion badge, and the vocabulary for every other day on this list.

The honest bit. Anthropic states no technical background is needed, so this genuinely is the open door of the week. It is labelled beta, so the content can change. Sign in before you start, because the page saves progress to your account and there is nothing to attach a badge to otherwise.

Day 2: LangChain, LangChain Essentials Python

Open the course · 18 lessons · 1 hour of video · free, certificate of completion, Python needed

What it is. Your first actual agent. You build one with create_agent, then add tools, MCP, memory, structured output and a human-in-the-loop step. LangChain is the framework most job ads mean when they say agent experience.

What it actually teaches. The difference between calling a model and running an agent: giving it tools, letting it decide, keeping state between turns, and putting a human approval step in the middle so it cannot run away with something.

What you walk away with. A certificate of completion issued by LangChain Academy, naming the course, with an issue date and a certificate ID on it. It has no expiry.

The honest bit. This is a developer course with Python examples. Do not confuse it with LangChain's paid Certified Agent Engineer exam, which is a different, paid product on the same site.

Day 3: Anthropic, Claude Platform 101

Open the course · 13 lessons · 1.5 hr · free, completion badge, coding required

What it is. The course where you stop using Claude as a chat window and start using it as something you build on. Your first API call, then building the agent loop by hand so you understand what every framework is doing for you.

What it actually teaches. The agent loop from scratch, tool use, built-in tools, Skills and MCP servers. Building the loop by hand once is the single best way to stop finding agents mysterious.

What you walk away with. A completion badge, and the ability to read any agent framework's docs and recognise what it is wrapping.

Day 4: Anthropic, Model Context Protocol: Advanced topics

Open the course · 11 lessons · 1.5 hr · free, completion badge, coding required

What it is. The hardest course on the list, and the reason it sits in the middle of the week rather than at the start. MCP is the standard agents use to reach tools and data, and this is the course about what is actually happening underneath.

What it actually teaches. Sampling, notifications, roots, and how MCP messages flow between client and server. The page carries interactive walkthroughs of each protocol flow, which is the clearest explanation of MCP available anywhere for free.

What you walk away with. A completion badge, and the ability to debug an MCP connection instead of restarting it and hoping.

The honest bit. Stated prerequisites are "Basic understanding of MCP servers and clients" and "Familiarity with async programming patterns". This is not a day-one course. If MCP is new to you, do day 3 first and give this one a second evening.

Day 5: LangChain, LangSmith Fleet, Build No-Code Agents

Open the course · 9 lessons · 1.5 hours of video · free, certificate of completion, explicitly no-code

What it is. The breather in the middle of the week, and the day that proves you do not have to write Python to build a working agent. LangChain's own words: "Start creating agents using everyday language. Learn how to build no-code agents for real work with LangSmith Fleet."

What it actually teaches. Agent building fundamentals, human-in-the-loop workflows, configuring what an agent is allowed to do, MCP integration, ambient agents that run quietly in the background, sub-agents, and computer use inside a sandbox. It covers the same ideas as days 3 and 4, from the other side of the keyboard.

What you walk away with. A LangChain Academy certificate of completion, and an agent you built by describing it rather than coding it.

The honest bit. Same certificate trick as day 2: nothing on the course page mentions a certificate, so finish it, exit the course, then go back in and the download button appears in the left-hand side menu. Do days 3 and 4 first if you can, because this lands better once you know what the agent loop and MCP actually are.

Day 6: Microsoft, Develop an agent with integrated tools

Open the credential · 45 minute timed lab · Intermediate · Microsoft publishes no price, coding required

What it is. The only hands-on lab credential in the week, and the hardest thing on the list. It is not a quiz. It is a timed lab in a real environment, and passing it puts a Microsoft credential on your Learn profile.

What it actually makes you do. Three things, in Microsoft's own words: "Create and configure agents by using the Microsoft Foundry SDK", "Configure a built-in agent toolset", "Create a custom tool".

What you walk away with. A Microsoft credential earned by doing rather than watching, and a genuine talking point in an interview, because creating a custom tool is the thing that separates people who have used agents from people who have built for them.

The honest bit. Two things. First, Microsoft never publishes a price for this one, in either direction, so check the credential page before you plan around it. Second, this is SDK work, so it needs real coding, and the study guide shows the real tasks: deploying a model and creating a custom tool that uses Azure Functions. Microsoft states candidates "should have experience creating agents by using Microsoft Foundry", so work through the study guide on a different day from the lab. Read the lab rules below before you start, because they are strict.

Day 7: MongoDB, Building AI Agents with MongoDB

Read the badge requirements · open the course (needs a sign-in) · 75 minutes of training plus a 10-question skill check · free Credly badge, Foundational

What it is. The finish line, and the only item in the week that ends in a Credly badge. That matters more than it sounds. Credly is the format LinkedIn reads natively, so this one lands on your profile as a verified credential with its own URL rather than a line of text you typed yourself.

What it actually teaches. The course page will not load until you sign in, so read the Credly badge page instead: it carries the requirements word for word. "Complete 75-minutes of training on how to build and deploy AI agents using MongoDB. Understand orchestration, data storage strategies, and how to integrate AI models with real-time databases." So it is the memory-and-data half of agent building, which is the part days 1 to 6 mostly skip.

What you walk away with. A verifiable Credly badge titled "Building AI Agents with MongoDB", tagged with AI Agents, GenAI and NoSQL Databases, plus an understanding of where an agent's memory actually lives.

The honest bit. There is a 10-question skill check at the end, so it is earned rather than handed over. Everything verifiable about this one comes off the Credly badge page, because the course itself is behind a sign-in. It is Foundational level and assumes you can follow code, though you are not building from scratch. A free MongoDB University account is required, and you will want a Credly account on the same email address so the badge lands properly.

Before you open the Microsoft lab

This lab has real rules, and the most common way people fail is not knowing them. All of it comes from Microsoft's own assessment lab FAQ.

1. ⏳ You get 30 minutes for a beginner lab and 45 for an intermediate one. Day 6 is intermediate, so you have 45 minutes.
2. 📋 Most labs contain 12 to 16 tasks. Read the whole task list first and do the ones you are sure of before the clock gets tight.
3. 🚫 You cannot pause and come back. In Microsoft's words, "You will NOT be able to save your work and come back later, so plan to complete all the tasks in a single session." Scoring begins in the background as soon as you leave the lab.
4. ⚠️ A lab that ends badly still counts. "Note that a cancelled lab or one that ends abruptly for any reason will be scored." Do not open one just to have a look.
5. 📆 You wait 72 hours before you can launch it again. That is the rule after every attempt, successful or not, so a careless practice run costs you three days.
6. 🙋 You can ask for a little extra time. When under 10 minutes remain, a pop-up offers an extension. Click yes.
7. 💻 Practical setup. Third-party cookies must be enabled, and use a desktop or laptop, because Microsoft recommends against a phone or tablet. Results can take up to 24 hours to appear in the Credentials tab of your Learn profile.

If you don't code, or don't have the week

Making them count

1. 📌 List them as you go, not at the end. Anthropic and LangChain go under LinkedIn's Licenses and certifications with the course name exactly as written on the page. The Microsoft one lands on your Microsoft Learn profile under Credentials. The MongoDB one arrives in Credly and pushes to LinkedIn in one click.
2. 🗣 Be accurate about the type. Anthropic issues completion badges. LangChain issues a certificate of completion. Microsoft Applied Skills is a credential earned through a hands-on lab, which is a genuinely stronger thing. MongoDB issues a verifiable Credly badge with its own URL. Saying that precisely costs nothing, and it is exactly the kind of accuracy that reads as competence in an interview.
3. 🔨 Keep the thing you built. Day 2 leaves you with a working agent, day 5 leaves you with a no-code one, and day 6 leaves you with a custom tool. Record thirty seconds of any of them running. A credential plus a demo is a completely different conversation from a credential on its own.