GitHub certifies agentic AI developers

GitHub and Microsoft have introduced a new certification aimed at developers and platform teams working with autonomous AI systems inside software delivery workflows. The credential, GitHub Certified: Agentic AI Developer, went live in beta on May 13 through Microsoft Learn and is tied to Exam GH-600. The program is designed around operating, integrating, supervising and governing AI agents in production-grade software development lifecycle workflows, with GitHub positioned as the system of record and control plane. Microsoft said the beta exam is intended for people already working with GitHub Copilot, agent workflows and AI-enabled development processes. ### What exactly is GitHub certifying here? Exam GH-600 is not framed as a coding test alone. Microsoft Learn says the certification validates expertise in deploying, operating, integrating and governing AI agents in production SDLC workflows, with an emphasis on reliability, safety and speed using GitHub as the control plane. The role description says candidates are expected to operate agent workflows, supervise autonomous behavior with GitHub controls, evaluate and tune outputs, configure custom agents and coordinate multi-agent execution safely. (techcommunity.microsoft.com) The May 13 announcement described the credential as role-based rather than feature-based. Microsoft said it is aimed at software developers, platform engineers, DevOps engineers, security engineers and technical product managers working with AI-assisted or agent-driven development workflows. ### Which skills show up on the syllabus? Microsoft’s announcement lists several of the exam’s core topics in operational terms. (learn.microsoft.com) The beta covers integrating agents into the SDLC, defining boundaries between planning, reasoning and execution, configuring tools, permissions and environments, managing memory, state and long-running execution, evaluating and improving agent performance, coordinating multi-agent workflows, and implementing guardrails and human-in-the-loop systems. (techcommunity.microsoft.com) The official exam page breaks the test into domains that include preparing agent architecture and SDLC processes, implementing tool use and environment interaction, and governing agent behavior inside GitHub-driven workflows. The associated study guide says candidates should already understand GitHub workflows, controls, code quality, security and review practices, and should have experience with coding agents, MCP servers, custom instructions, custom agents and tools. (techcommunity.microsoft.com) A related Microsoft Learn module sets out the underlying model GitHub wants candidates to know: plan, act and evaluate. That training also points to traceability requirements, risks, anti-patterns and contributor review practices for agent-generated work. ### Why does GitHub keep calling itself the control plane? Microsoft Learn repeatedly describes GitHub as the “system of record and control plane” for agent activity. (learn.microsoft.com) In the training materials, that means using repository workflows, checks, reviews, branch protections and other governance controls to supervise what agents can do and how their work is validated. The course materials tie that model to safe, traceable workflows rather than autonomous execution without oversight. (learn.microsoft.com) The same training path links agent use to execution boundaries, permissions, MCP servers and environment controls. Those details place the certification closer to production operations and governance than to prompt-writing alone. ### Who is this beta exam for in practice? GitHub’s published audience profile says the certification is for people who already have SDLC experience and familiarity with GitHub controls, reviews and security practices. (learn.microsoft.com) The role description also says candidates work with architects, platform engineers, DevOps engineers, application developers, product managers and security engineers to develop, deploy and manage agents inside the GitHub platform. (learn.microsoft.com) That positioning matters because the exam assumes production exposure. Microsoft’s announcement says the test is about operating AI systems safely and effectively in real-world development workflows, not simply knowing a single product feature. ### What does the beta process look like? The exam page says GH-600 is currently in beta, lasts 120 minutes and is proctored. (learn.microsoft.com) Microsoft Learn also says candidates will not receive results immediately, with scores scheduled to be released about eight weeks after the beta period ends. A passing score is 700 or higher, according to the study guide. Microsoft Reactor has scheduled a livestream on May 28 focused on the GH-600 certification. (techcommunity.microsoft.com) The event page says the first 100 people who take the beta exam on or before May 31 can receive 80% off market price, and names Ari LiVigni, a senior learning advocate at GitHub, as the speaker. ### Where can candidates go next? Microsoft Learn has posted the certification page, the GH-600 study guide and a training path called “Developing in Agentic AI Systems” in two parts for candidates preparing for the exam. (learn.microsoft.com) The beta exam can be scheduled through the certification page now, and the Microsoft Reactor walkthrough is set for May 28, 2026. (developer.microsoft.com)