GitHub Certified: Agentic AI Developer (GH-600)

Move from coding assistance to governed agent workflows

AI coding agents can plan work, modify code, run tools and raise pull requests, but developers must stay in control of what reaches production. This one-day course is for experienced engineers who operate and supervise agents within GitHub-based software development. It explores how to introduce autonomous work into established delivery practices while preserving review, traceability and accountability.

Design agents around the software lifecycle

Learn to define agent responsibilities, inputs, outputs and measurable success criteria. Explore how to separate planning from execution, choose suitable autonomy levels and set human approval points. GitHub acts as the system of record and control plane, with repository boundaries and branch-based workflows helping teams examine each change.

Connect tools and preserve context

Examine agent tools, Model Context Protocol (MCP) servers, permissions and execution environments. Consider how agents use short- and long-term memory, persist state and recover from context drift. Explore safe retries, escalation and rollback when an agent encounters errors or cannot complete a task.

Evaluate and coordinate agent activity

Review how logs, scans, plans and pull-request artefacts support evaluation and root-cause analysis. Explore tuning instructions, memory and tool access using evidence from failed or degraded runs. Consider multi-agent orchestration, isolation, hand-offs and conflict resolution when agents work on related tasks.

Protect quality without blocking delivery

Understand how least privilege, policy controls, explicit authorisation and human-in-the-loop workflows keep agent actions within agreed limits. Training follows the six GH-600 exam domains and supports certification preparation. Further practice with GitHub Copilot and live development environments may be useful before the assessment.