GH-300 Study Hub
A study hub for GH-300: GitHub Copilot, organized around the six official exam domains rather than a course timetable — because that’s how the exam is actually scored, and how you should actually revise.
Every fact on this site is checked against official sources. Domain names and weightings come from the Microsoft Learn skills outline; product behavior comes from GitHub's own documentation. Anything that could not be confirmed against an official source has been left out rather than guessed at.
Microsoft Learn — Study guide for Exam GH-300 · skills measured as of 7 August 2026. Built and verified 26 August 2026.
The exam#
| Delivery | Online proctored, through Pearson VUE |
| Seat time | 120 minutes (100 minutes of exam, plus check-in and instructions) |
| Level | Intermediate. No formal prerequisites |
| Languages | English, Spanish, Portuguese (Brazil), Korean, Japanese |
| Microsoft Learn during the exam | Not available — GitHub exams are closed-book |
| Retakes | 24 hours after a first failure, then 14 days between attempts, max 5 in any 12 months |
| Renewal | Expires annually. Renewal is free, unproctored and open-book on Microsoft Learn; the window opens 6 months before expiry |
| Price | Varies by country and region — shown during Pearson VUE scheduling |
Question count and price
Microsoft does not publish a question count for GH-300, and states only that pricing is "based on the country or region in which the exam is proctored". General Microsoft guidance is that most exams contain 40–60 questions, "however, the number can vary". You'll see the exact price for your region during Pearson VUE scheduling.
The six domains#
Weightings are official. Note in particular that responsible AI is the second-largest domain on the exam — worth planning your revision around.
Use GitHub Copilot responsibly
Risks and limitations of generative AI, ethical use, identifying harms and mitigations, validating AI output.
15–20% Domain 2Use GitHub Copilot features
Copilot in the IDE, Copilot CLI, agent mode, Copilot Edits, MCP, code review, Spaces, and organization-wide policy management.
25–30% Domain 3Understand Copilot data and architecture
Data usage and flow, input processing and prompt building, proxy filtering, the suggestion lifecycle, and LLM limitations.
10–15% Domain 4Apply prompt engineering and context crafting
Prompt structure and context, how context is determined, zero-shot and few-shot prompting, prompt process flow and chat history.
10–15% Domain 5Improve developer productivity
Code generation, refactoring and documentation, reducing context switching, modernizing legacy code, tests, edge cases, security and performance.
10–15% Domain 6Configure privacy, content exclusions, and safeguards
Content exclusions and editor settings, ownership and limitations of outputs, public code matching, and troubleshooting.
10–15%Where to spend your time
Domains 1 and 2 together account for 40–50% of the exam. If your revision time is limited, that is where it belongs. The remaining four domains are evenly weighted at 10–15% each, so there is no single "safe to skip" area.
Study material#
Labs
Eight practical walkthroughs — surfaces, prompting, workflows, agent mode and MCP, enterprise configuration, security testing, customization, and a capstone.
PracticeQuestion bank
Practice questions grouped by official domain, each with an explanation and a cited source.
RevisionCheatsheets
Condensed reference cards, including printable pocket formats for last-minute revision.
Deep divePlans and AI Credits billing
Plans, feature availability by tier, and how usage-based AI Credits billing works.
OfficialResources
Official GitHub and Microsoft study materials, the free practice assessment, video courses, and registration.
Facts worth committing to memory#
These are the specifics most likely to be tested, current as of August 2026.
| Topic | Current answer |
|---|---|
| Chat modes | Ask, Plan, and Agent. Only agent mode runs terminal commands, and only with your approval |
| Extensibility | MCP servers, custom agents, agent skills, and third-party agents |
| Billing | GitHub AI Credits, usage-based since 1 June 2026. 1 credit = $0.01. Code completions are not billed in credits |
| Plans | Free, Student, Pro, Pro+, Max, Business, Enterprise |
| Models | GPT-5.3-Codex is the base and LTS model, designated 18 March 2026 |
| Workspace context | #project |
| Custom agents | .github/agents/NAME.md |
| MCP config | .vscode/mcp.json — the primary, shareable location |
| Policy scope | Enterprise → organization. There is no team scope |
| Content exclusions | Repository, organization, or enterprise. Up to 30 minutes to propagate |
Each domain page covers these in context, with a citation to the official documentation behind every claim.