In short: Roblox is upgrading its built-in AI assistant with agentic capabilities including a planning mode that analyses game code before proposing action plans, procedural 3D model generation, mesh generation, and self-correcting loops that test and refine outputs. The update also adds MCP client integration with third-party tools like Claude and Cursor, with a roadmap toward multi-agent parallel workflows in the cloud.
Roblox is advancing its AI assistant, transforming it from a simple code-suggestion tool into a more sophisticated junior development partner. This update enables the AI to plan, build, and test games by analyzing existing code and data models, asking clarifying questions, and proposing action plans that developers can approve before implementation.
What the New Tools Do
The new Planning Mode allows the assistant to serve as a collaborative planner. Instead of merely responding to prompts with code snippets, it examines a game’s existing code and data model and engages developers in a dialogue to create an editable action plan. This process empowers developers to review and modify the plan, ensuring it aligns with their vision before the assistant executes it.
Procedural Models, set to launch soon, will enable developers to create 3D objects defined by code rather than static meshes. For instance, a developer can instruct the assistant to generate a bookcase and then adjust its attributes—such as the number of shelves and material—through parameters rather than manual modeling. This approach ensures the generated objects understand physical relationships: a staircase recognizes its step height, and a table acknowledges that its legs support its surface.
Mesh Generation also enhances the AI's functionality by allowing it to place fully textured 3D objects directly into the game world through prompts. This builds on Roblox’s Cube foundation model, which was upgraded to include 4D generation, adding interactivity to objects so they function correctly in-game rather than remaining static. Early access data revealed that players using 4D generation experienced a 64% increase in playtime on average.
The Agentic Loop
A pivotal improvement is the self-correcting system. The assistant can now test various game aspects, identify issues, suggest solutions, and integrate those findings back into its planning process. Roblox refers to this as agentic loops, which are cycles of planning, execution, testing, and refinement requiring progressively less human oversight over time.
Roblox's development roadmap aims to further enhance this system by enabling multiple AI agents to collaborate in parallel, managing extensive workflows in the cloud rather than being confined to local studio sessions. The company is also integrating with third-party tools like Claude, Cursor, and Codex, incorporating a built-in MCP client into Roblox Studio’s assistant for external AI service connectivity.
The long-term vision, articulated since the Cube foundation model was open-sourced in March 2025, is for developers to describe a game in natural language, allowing the AI to generate all necessary assets, environments, code, animations, and interactive behaviors. The newly introduced agentic tools are incremental but signify a meaningful shift from AI acting as autocomplete to becoming a true collaborator.
The Vibe-Coding Parallel
This update coincides with a broader shift in software development practices. The rise of vibe coding—where developers describe desired outcomes in natural language for AI to generate code—has already led to an 84% increase in App Store submissions, prompting Apple to regulate low-quality AI-generated apps. Similar trends are emerging in game creation, indicating that barriers to building playable content are rapidly diminishing.
For Roblox, this presents both an opportunity and a challenge. Increased creator activity results in more games, driving platform engagement, but only if these games meet quality standards. The new planning mode and self-correcting loops address this challenge by guiding creators through structured processes, aiming to produce superior outputs rather than relying on initial AI-generated prompts.
Several third-party AI tools for Roblox game creation have surfaced, including Lemonade, SuperbulletAI, and BloxBot. By embedding agentic capabilities within Roblox Studio, the company aims to maintain the primary creation experience on its platform, reducing fragmentation with external tools.
The Business Context
Roblox’s investment in AI creation tools is supported by impressive commercial growth. Daily active users reached 144 million in Q4 2025, rising from 85 million the previous year, while monthly active users grew from 280 million to 380 million. The total revenue for 2025 was $4.9 billion, reflecting a 36% increase, with 2026 projections estimating $6 to $6.2 billion. Robux purchases hit $6.79 billion in 2025.
These figures are crucial as they determine Roblox’s capacity to invest in AI infrastructure and enhance its creator ecosystem through better tools. A platform with 380 million monthly users and nearly $5 billion in revenue can afford to develop foundational models, train agentic systems, and manage the computing demands of AI-assisted game creation at scale—an advantage that smaller platforms might lack.
The upcoming Roblox Developers Conference in September in San Jose is expected to showcase the next phase of this roadmap. The introduction of the agentic assistant positions Roblox as a leader in the transition from AI-assisted coding to AI-assisted product development, where the AI not only writes code but also plans, builds, tests, and refines its creations. The effectiveness of this new approach in producing quality games will be evaluated in the coming year.