What does it cost?
The GH-600 exam costs $165.
Exam fee
$165
GitHub prices the exam by the country or region where it is proctored; $165 is the price in the United States.
Source: GitHub exam page
GitHub exam
GH-600 Exam facts
GitHub Certified: Agentic AI Developer
- Minutes
- 120
- Exam fee
- $165
- Passing score
- 700
Independent: not affiliated with GitHub. These exam facts are free; the study bundle adds a study guide and two practice exams.
Short answer
GH-600 is the exam for the GitHub Certified: Agentic AI Developer certification, for developers and platform engineers who operate, constrain and audit AI agents on GitHub repositories. The exam runs 120 minutes, costs $165 and takes 700 of 1,000 to pass.
Sources: exam page, exam guide
GitHub's published facts for GH-600 on one page, from GitHub and Microsoft Learn. Last updated
Study bundle
Prepare for GH-600
$19
Buy now - $19
Secure checkout, in a new tab.
Your receipt has a license key. Enter it at overqual.com/download for the three PDFs, stamped with your email and order number. They are yours to keep.
Try 5 free sample questions first See 3 pages inside
- A 60-page study guide: every exam domain, weighted like the exam
- Two practice exams: 120 original questions, every answer explained
- Every bundle includes 6 months of access: download the PDFs, and any new editions, until 6 months after your purchase date.
Edition 1.0, October 2026
How many questions and how long?
The GH-600 exam runs 120 minutes. GitHub has not published the number of questions.
Source: GitHub exam page
The exam
- Questions
- Not published
- Time
- 120 minutes
- Languages
- English
What score do you need to pass?
The GH-600 passing score is 700, on a scale of 1-1,000.
1-699
700+ PASS
Source: GitHub exam guide
What does the exam cover?
The GH-600 exam covers six domains, weighted 15-20% / 20-25% / 10-15% / 15-20% / 15-20% / 10-15%.
Source: GitHub exam guide
D115-20%
Prepare agent architecture and SDLC processes
D220-25%
Implement tool use and environment interaction
D310-15%
Manage memory, state, and execution
D415-20%
Perform evaluation, error analysis, and tuning
D515-20%
Orchestrate multi-agent coordination
D610-15%
Implement guardrails and accountability
Prepare agent architecture and SDLC processes
- 1.1Integrate agents into the SDLC
- 1.2Define boundaries between planning, reasoning and action
- 1.3Configure observability and control for autonomous agents
Implement tool use and environment interaction
- 2.1Select and configure agent tools
- 2.2Configure MCP servers
- 2.3Integrate agents within development environments
- 2.4Operate agents with safe execution paths and robust error handling
Manage memory, state, and execution
- 3.1Agent memory strategies
- 3.2Persist state and manage context drift
- 3.3Continuity across tools and environments
Perform evaluation, error analysis, and tuning
- 4.1Success criteria and evaluation signals
- 4.2Analyze failures and root causes
- 4.3Tune behavior from evaluation results
Orchestrate multi-agent coordination
- 5.1Operate multi-agent workflows
- 5.2Observability for multi-agent behavior
- 5.3Detect and respond to failures and degraded behavior
- 5.4Manage agent lifecycle
Implement guardrails and accountability
- 6.1Define autonomy levels
- 6.2Implement guardrails and human-in-the-loop workflows
Inside the study bundle
Study guide chapters
Eleven chapters. Chapters 3 to 8 cover the six exam domains, weighted like the exam.
- 01 What the exam is p. 3
- 02 The GitHub agentic landscape p. 7
- 03 Agent architecture and the SDLC p. 13
- 04 Tools, MCP and the execution environment p. 18
- 05 Memory, state and execution p. 30
- 06 Evaluation, error analysis and tuning p. 35
- 07 Multi-agent coordination p. 40
- 08 Guardrails and accountability p. 45
- 09 Final review p. 51
- 10 Sources p. 58
- 11 Edition and updates p. 60
Look inside the study guide
Three pages of the Edition 1.0 study guide: the contents and the first two pages of a chapter.
What is free here, and what is in the bundle?
The exam facts and 5 sample questions are free on this site. The study guide chapters and the two practice exams are in the bundle.
| Item | Free on this site | In the bundle |
|---|---|---|
| Exam facts and change log | Yes, with links to the GitHub sources | Chapter 1 of the study guide sums up the exam's format and scoring. |
| 5 sample questions | Yes | No. None of the 5 appear in the practice exams. |
| Domain chapters and final review | No | Yes, in the study guide |
| Two practice exams with answer keys | No | Yes, 120 questions with an explanation for every answer |
Questions before you buy
Is this official?
No, the GH-600 bundle is not official. Overqual is independent and is not affiliated with, sponsored by, or endorsed by GitHub or Microsoft.
Does it include real exam questions?
No, the GH-600 bundle contains no exam questions, and every practice question is original.
How is the GH-600 bundle different from GitHub's free materials?
GitHub's free pages set out what the GH-600 exam covers. The GH-600 bundle adds a 60-page study guide that covers every exam domain, and two practice exams with 120 original questions, every answer explained. When GitHub revises the exam, a new edition follows if the changes warrant it, included while your access lasts.
How do I get updates?
Updates to the GH-600 bundle arrive as new editions. Every bundle includes 6 months of access: download the PDFs, and any new editions, until 6 months after your purchase date. Download each new edition at overqual.com/download with the license key from your receipt email. The change log on the exam facts page says what changed.
Can I get a refund?
No. Sales of the GH-600 bundle are final, because downloaded PDFs cannot be returned. Read the free exam facts and try the 5 free sample questions first. If a download fails or you paid twice by mistake, contact Overqual to have the download fixed or the second payment refunded.
How is GH-600 different from GH-300?
GH-600 (GitHub Certified: Agentic AI Developer) tests how you operate, supervise and govern AI agents that act on a repository: tool permissions, MCP servers, memory and state, failure analysis, multi-agent coordination and guardrails. GH-300 (GitHub Copilot) covers using Copilot responsibly, its features, data and architecture, prompt engineering, developer productivity, and privacy, content exclusions and safeguards; it runs 100 minutes and is offered in five languages, where GH-600 runs 120 minutes in English only. Neither certification page lists a prerequisite or a retirement date. Sources: GH-600 certification page, GH-300 certification page.
Does GH-600 include labs or hands-on tasks?
GH-600 is proctored and, in the certification page's words, "You may have interactive components to complete as part of this exam." Microsoft does not say which formats a given exam uses before you sit it; its exam sandbox shows the interface, and its exam experience page demonstrates multiple choice, drag and drop, build list, hot area, active screen, case studies and labs. Microsoft Learn cannot be opened during GitHub exams. Sources: certification page, exam duration and exam experience.
Is GH-600 still a beta exam?
GH-600 no longer carries a beta label. Microsoft announced Exam GH-600 (beta) on its Skills Hub Blog on May 13, 2026, with general availability planned for July 2026, and a Microsoft Reactor session on May 28, 2026 walked through Microsoft's GH-600 study guide for the beta exam period. The certification page today lists the exam with no beta label, at 120 minutes in English. No Microsoft page gives the date the beta ended. Sources: Microsoft Skills Hub Blog announcement, Microsoft Reactor event page, certification page.
Ready for GH-600?
Three PDFs, 106 pages: the study guide and two practice exams.
Which parts of GitHub does the exam expect you to configure?
GitHub is the system of record and control plane throughout the GH-600 outline. The outline's most frequent verbs are "configure" and "implement", so the pieces below are things to set up and read, not names to recognize: where agents run, what reaches them, what stops them and what they leave behind.
Agent surfaces
- GitHub Copilot cloud agent
- Copilot CLI
- Agent mode in Visual Studio Code
- Custom agents (.agent.md files in .github/agents)
- Copilot SDK custom agents and sub-agents
Tools and environment
- Built-in tools (read, edit, search, execute) and the tools allowlist
- MCP servers, the GitHub remote MCP server, MCP registries and allow lists
- copilot-setup-steps.yml and Agents secrets (COPILOT_MCP_ prefix)
- The agent firewall and its custom allowlist
- GitHub Actions workflows that invoke an agent
Controls that enforce
- Rulesets and branch protection
- CODEOWNERS and required reviews
- Approval of workflow runs on agent pull requests
- Hooks (preToolUse deny)
- Organization and enterprise Copilot policies
Evidence and state
- Session logs
- Copilot Memory (repository facts and user preferences)
- Durable artifacts: plans, issues and pull request descriptions
- Audit log entries for agent actions
- CodeQL, secret scanning and Copilot code review as evaluation signals
How is it delivered?
The GH-600 exam is proctored, scheduled through Pearson VUE.
- Delivery
- Proctored, scheduled through Pearson VUE
- Languages
- English
- Validity
- 2 years
- Retakes
- 24 hours after the first attempt; for subsequent retakes the amount of time varies
- Test aids
- Microsoft Learn access is not available on GitHub exams
Source: GitHub exam page
Fact sheet
Show the 27 factsHide the facts
- Time
- 120 minutes Source: GitHub exam page
- Fee
- $165 Source: GitHub exam page
- Passing score
- 700 Source: GitHub exam guide
- Score range
- 1-1,000 Source: Microsoft Learn scoring page
- Languages
- English Source: GitHub exam page
- Delivery
- Proctored, scheduled through Pearson VUE Source: GitHub exam page
- Validity
- 2 years Source: GitHub exam page
- Retakes
- 24 hours after the first attempt; for subsequent retakes the amount of time varies Source: GitHub exam page
- Test aids
- Microsoft Learn access is not available on GitHub exams Source: Microsoft Learn exam experience page
- Domain weights
- D1 15-20% / D2 20-25% / D3 10-15% / D4 15-20% / D5 15-20% / D6 10-15% Source: GitHub exam guide
- Level on Microsoft Learn
- Intermediate Source: GitHub exam page
- Exam sandbox
- Demo of the exam interface at ghcertdemo.starttest.com Source: GitHub exam page
- Renewal
- GitHub is transitioning to Microsoft's recertification process; certifications that expire before it is available are extended by 6 months Source: GitHub exam page
- Exam account
- Microsoft recommends registering with a personal Microsoft account; records under a work or school account are lost if you leave it Source: GitHub exam page
- Roles
- AI Engineer, App Maker, Data Engineer, Developer, DevOps Engineer, Solution Architect Source: GitHub exam page
- Certification earned
- GitHub Certified: Agentic AI Developer Source: GitHub exam page
- Outline size
- 6 domains, 19 skill groups, 65 bullets Source: GitHub exam guide
- Preview features
- Most questions cover GA features; preview features may appear if commonly used Source: GitHub exam guide
- Official training
- Microsoft Learn modules: Foundations of Agentic AI in GitHub; Designing Agent Architecture and SDLC Integration; Tooling, MCP, and Agent Execution Environments Source: GitHub exam guide
- Course
- GH-600T00-A: Developing in Agentic AI Systems, 1 day Source: Microsoft Learn course page
- Documentation links
- One GitHub Docs link per domain, listed in Microsoft's GH-600 study guide Source: GitHub exam guide
- Practice assessment
- None listed on the certification page Source: GitHub exam page
- Extra time if the exam is not in your language
- 30 minutes on request (GH-600 is offered in English only) Source: GitHub exam guide
- Breaks
- 5 minutes are built into the exam time for unscheduled breaks; the clock keeps running Source: Microsoft Learn exam experience page
- Scoring
- No penalty for guessing; some questions may not be included in your score Source: Microsoft Learn scoring page
- Accommodations
- Available on request Source: GitHub exam page
- Question types on the exam experience page
- Multiple choice, drag and drop, build list, hot area, active screen, case studies and labs Source: Microsoft Learn exam experience page
What has changed, and when?
Changes to the exam and to the products it covers, newest first.
- Microsoft announced the GitHub Certified: Agentic AI Developer certification and the GH-600 beta, with general availability planned for July 2026. Microsoft Tech Community
To watch GitHub directly, check the exam page.