LangChain Certified Agent Engineer

Six full-length practice exams — 240 questions in total — built to mirror the LangChain Certified Agent Engineer (LCAE) certification.

Matches the real exam structure

Every test follows the official blueprint: 40 questions spread across the four sections of the Agent Development Lifecycle — Build, Test, Deploy and Monitor — weighted at 25% each, on a 120-minute timer, exactly like the real thing. You will meet both single-answer and select-all-that-apply questions, because the certification uses both.

Written from the official documentation

Every question is grounded in the current LangChain, LangGraph and LangSmith documentation, and every explanation names the page its answer comes from. A wrong answer sends you to the exact section you need to re-read, rather than leaving you to guess which topic let you down.

Includes hands-on LangSmith questions

The real certification expects you to open LangSmith and read a live dashboard, and these tests do the same. A free companion dataset loads into your own LangSmith workspace in about ten minutes, and a few Monitoring questions in each test send you there to find the answer: which account has the highest error rate, where the token spend is going, which experiment scored better. This is the part most candidates are least prepared for, and the only way to get comfortable with it is to practise it.

Set up the LangSmith dataset (before you start)

The dataset isn't inside the course. You load it into your own free LangSmith workspace with a few short scripts. To find them, search GitHub for MBehtemam lcae-practice-exam-data, or use the direct link in the course welcome message.

Clone the repo, add your LangSmith API key to the .env file, run uv sync, then run the reset, upload_traces and upload_experiments scripts in that order. You'll get a project called lcae-udemy-practice (about 225 traces) and a dataset called support-faq-eval with two experiments. Run the upload right before you start a test, because the traces are timestamped relative to the upload time, and questions about "the last 24 hours" give different answers if you wait more than about 6 hours.

What you should already know

These are practice exams, not a course. You will get the most from them once you have worked through the LangChain Academy material or have equivalent hands-on experience: building agents with create_agent or deepagents, writing and running evaluations, and deploying to LangSmith Deployment.

How to use them

Sit each test under real conditions — closed notes, one sitting, full timer. Then review every explanation, including for the questions you answered correctly. Retake a test only after reading the documentation behind whatever you missed, so you are learning the material rather than memorising the answers.