Roblox is rolling out agentic enhancements to Roblox Studio and Roblox Assistant to speed up planning, building, and testing for creators, leveraging AI and agentic workflows.

The system can request specifics such as visual style and asset creation methods to align with a creator’s intent and refine plans iteratively.

The overarching goal is to close the gap between creative vision and execution by enabling agentic loops that improve accuracy over time.

Planning Mode works with playtesting tools to read logs, capture screenshots, simulate inputs, identify bugs, and feed feedback for automatic fixes.

Creators like Malt of Solo Hunters point to potential future capabilities such as overnight bug fixes and pull-request-based integration.

Mesh generation will add fully textured 3D objects via prompts, leveraging Roblox’s Cube foundation model and moving toward 4D generation for interactive behavior, which has shown to boost playtime among early adopters.

Procedural models will let developers generate parametric 3D objects via prompts, with attributes like dimensions and materials adjustable without manual modeling, and objects understanding physical relationships.

Improved Planning Mode enables multi-step, collaborative development by having AI analyze code and data, ask clarifying questions, and convert prompts into a detailed, editable action plan.

This mode allows the AI to review a game’s code and data model, pose clarifying questions, and deliver an actionable plan for the developer to approve prior to execution.

Long-term, the vision is for developers to describe a game in natural language and have AI generate assets, environments, code, animations, and interactivity, with agentic tools acting as incremental steps toward AI-as-collaborator.

These features build on the fact that a sizable share of top creators—about 44%—are already using AI tools to plan, build, and test, with ongoing improvements and a roadmap ahead.

Planning Mode turns Assistant into a collaborative partner that analyzes a game’s code and data models, asks clarifying questions, and produces an editable action plan before changes are made.