This course contains the use of artificial intelligence.

These practice exams are unofficial and independently written. They are not affiliated with, endorsed or sponsored by GitHub, Inc. or Microsoft Corporation. GitHub, GitHub Copilot and the GitHub Certified: Agentic AI Developer credential are trademarks of GitHub, Inc.; Microsoft and Microsoft Learn are trademarks of Microsoft Corporation. The names are used only to say which exam these tests help you prepare for. No question comes from a real GH-600 exam.

Exam GH-600 (GitHub Certified: Agentic AI Developer) checks whether you can run AI agents safely inside a real software delivery workflow, with GitHub as the system of record. The questions are about decisions: what an agent should be allowed to do, which tools and MCP servers it gets, where its memory and state live, how you tell a good agent pull request from a bad one, how several agents work in parallel without colliding, and where a human has to stay in the loop. These two practice exams train exactly those decisions.

What you get

• Two full practice exams of 60 questions each (120 in total), 120 minutes each, the same time the real exam allows
• Questions weighted to Microsoft's published GH-600 skills outline across all six domains: agent architecture and SDLC, tool use and MCP, memory and state, evaluation and tuning, multi-agent coordination, and guardrails
• About 80% single-answer questions and 20% "choose two" or "choose three" questions
• An explanation for every answer option, plus an overall explanation that names the skill being tested and links the GitHub Docs page it is based on
• Scenario questions where the right answer depends on knowing how GitHub Copilot agents actually behave, not on spotting the longest option

Topics you will practice

• Copilot cloud agent: branches, draft pull requests, workflow approval, session logs and steering
• Custom agents: agent profiles, tool lists, MCP servers per agent, versioning, testing and release
• MCP: repository configuration, the GitHub MCP server, Agents secrets, registries and enterprise allow lists
• Copilot setup steps, runners, the agent firewall and cloud sandboxes
• Custom instructions, agent skills, hooks and Copilot Memory
• Copilot CLI tool permissions, autopilot, plan mode, subagents and /fleet
• Rulesets, CODEOWNERS, audit logs, automations and least-privilege guardrails

How to use these tests

Take one exam under timed conditions, then read the explanation for every question, including the ones you got right. The pass mark in this course is 70%, set as a study target. Microsoft reports GH-600 results on a scaled score where 700 is a pass, which is not a percentage, so a practice score is a study signal only and does not predict your exam result. The real exam may also contain interactive item types that a multiple-choice practice test cannot reproduce.

GitHub features change often. The questions were checked against GitHub Docs on 12 November 2026; please check Microsoft Learn and GitHub Docs for the current exam details and feature names before you sit the exam.