Unreal Engine 5.9 MegaJam Preview Arrives with Agent-Built Features

Epic Games has made an Unreal Engine 5.9 MegaJam Preview available ahead of its November 12–15 AI MegaJam. The November 9 announcement is notable both for what the preview contains and for how Epic says it was assembled. Rather than a typical months-long preview process, the company says a small group of engineers directed parallel AI coding agents to bring tools from its UE6 development branch into UE5 and to revive browser deployment as a target.

Epic describes the preview as its first Unreal Engine release developed primarily using agentic workflows. Human engineers directed, reviewed and tested the work, while agents contributed code. According to the announcement, the web platform work was derived in part from the older UE4 HTML5 platform. This is a description of Epic's development method, not evidence that every resulting change is as mature as a conventional release.

The preview expands Unreal MCP, an experimental bridge through which agents can interact with editor systems. It also introduces an experimental Unreal Assistant experience for external coding agents, broader Python scripting interfaces and an Unreal Canvas content-creation workflow. The browser target runs through WebAssembly and WebGPU. These are separate tools with different risks; their presence in one build does not make them universally production-ready.

Epic is unusually explicit about the limitations. It says the build received roughly a week of focused development and a shorter quality-assurance pass than normal. Some agent-authored code still needs cleanup, the web renderer contains prototypes, and the full UE5.9 release is expected later. The forum announcement includes a detailed list of known failures affecting web packaging, browser memory, editor scripting and toolset availability.

For a studio interested in AI-assisted game production, the preview offers a useful early look at workflows that may eventually shorten iteration. A sensible experiment would copy a small C++ prototype project, enable one AI toolset at a time and record which editor operations still require manual intervention. Keep the original production branch untouched and log crashes or missing output separately from agent suggestions.

The wider significance is not that AI has replaced engine engineers. It is that Epic is publicly testing a development model in which several agents can implement work under human supervision, while exposing more of the engine to external tools. Whether that translates into dependable production workflows will depend on later testing and stable releases.