Exam GH-600
Agentic AI Developer Trainer
Independent preparation for GitHub Certified: Agentic AI Developer, grounded in the current Microsoft Learn outline and GitHub documentation.
Official certification artwork shown for reference.
Target certification badge shown for reference. Independent third-party trainer; not affiliated with or endorsed by GitHub.
Guided progression
Loading path
Exam simulation
GH-600 Exam mode
Exam mode builds a 60-question practice simulation from all six domains. Its distribution stays inside the current official ranges: 17%, 23%, 12%, 17%, 18%, and 13%.
The official certification page gives 120 minutes for the assessment. This practice timer starts after you press Start Exam. The 70% practice result is not Microsoft's scaled 700 score.
Official source
Official GH-600 study guide
This compact viewer follows the Microsoft Learn outline last updated 13 May 2026. Open the official source for current details, updates, and linked GitHub documentation.
Skills measured
- Agent architecture and SDLC processes: 15-20%
- Tool use and environment interaction: 20-25%
- Memory, state, and execution: 10-15%
- Evaluation, error analysis, and tuning: 15-20%
- Multi-agent coordination: 15-20%
- Guardrails and accountability: 10-15%
Exam profile
- Credential: GitHub Certified: Agentic AI Developer
- Level: Intermediate
- Products and subjects: GitHub, application development, and AI
- Official assessment time: 120 minutes
- Study guide checked: 26 August 2026; source updated 13 May 2026
Core preparation focus
- Operate, integrate, supervise, and govern production agents with GitHub as the system of record and control plane
- Use SDLC controls, GitHub workflows, code review, security, GitHub Copilot, MCP, custom instructions, custom agents, tools, and setup steps
- Expect mostly generally available features; commonly used preview features can appear
Architecture and SDLC
- Select suitable SDLC work, avoid agent anti-patterns, and define inputs, outputs, constraints, and success criteria
- Separate planning from action, produce a structured plan, validate it, and gate execution until required checks pass
- Set autonomy from risk, create inspectable GitHub artifacts, and add human intervention where judgment reduces material risk
Tools, environments, memory, and state
- Select and configure only required tools and permissions; add MCP servers, the GitHub remote MCP server, registries, and allowlists
- Scope execution by repository, branch, CI workflow, runtime constraints, and controlled branch or pull-request creation
- Implement explicit errors, bounded retries, rollback, escalation, and attribution
- Choose memory type and lifecycle, persist valid progress, resume safely, correct drift, and remove conflicting or stale context
Evaluation, coordination, and accountability
- Define expected outcomes and qualitative and quantitative signals; use scanners, logs, plans, traces, outputs, and artifacts
- Classify reasoning, tool, context, and environment faults before tuning instructions, memory, or tool access
- Coordinate and isolate multiple agents, reconcile conflicts, retain decisions and handoffs, detect degradation, recover, and manage agent lifecycle
- Classify action risk, enforce least privilege and policy, require explicit authorization for sensitive work, and remove approvals that add no material safety