Roblox's AI Assistant Gains New Agentic Tools to Plan, Build, and Test Games

Roblox’s AI Assistant Gains New Agentic Tools to Plan, Build, and Test Games

3 Min Read

Roblox is unveiling new features to assist developers in planning, building, and testing games on its platform, as revealed exclusively to TechCrunch. The company is enhancing Roblox Assistant, an AI tool for game development, aiming to support creators throughout the entire process.

AI tools that offer solutions in a single step might not always align with a creator’s vision. Hence, the introduction of an improved “Planning Mode” transforms Assistant into a collaborative partner, capable of analyzing a game’s code and data model, asking clarifying questions, and forming prompts into actionable plans.

Planning Mode allows developers to draft a game plan, refine it with feedback, finalize the approach, and implement it. Creators can adjust the plan and add context to ensure their intentions are reflected accurately before any changes occur.

For instance, creators might instruct Assistant to “create a park mini game with a fountain and foliage where characters collect coins.” Assistant might then inquire about the visual style, offering choices like cartoony, realistic, and fantasy, and ask how the park’s assets should be generated, providing options such as building from scratch or using Creator Store models.

Once a plan is established, Planning Mode utilizes Roblox’s other AI tools in creating the game, including two newly announced tools: Mesh and Procedural Model Generation, designed to expedite development.

Mesh Generation simplifies adding textured 3D objects into the game, replacing low-quality placeholders. During initial development, developers usually create temporary assets to understand player interaction. Mesh Generation allows creators to quickly produce 3D models, replacing those placeholders.

Creators can ask Assistant to generate a campfire, add light for realism, and set the scene at night.

Roblox will soon introduce “Procedural Models” to enable developers to create editable 3D models with code and Assistant. Assistant’s understanding of 3D space and physical relationships lets creators use prompts to place and scale objects relative to each other.

Attributes like the number of shelves in a bookcase or the height of a staircase can be adjusted dynamically, producing editable building blocks for refinement and reuse.

Nick Tornow, Senior Vice President of Engineering, stated, “The launch of our agentic features in Roblox Studio diminishes barriers between creative vision and execution. Creating with Planning Mode and our Procedural Generation tools empowers creators to transform their concepts into gameplay. Assistant functions as a development partner, accelerating planning, building, and testing, helping creators move from idea to reality faster.”

As Planning Mode executes a plan, it will employ playtesting tools to examine output logs, capture screenshots, utilize inputs like a keyboard and mouse to review design and gameplay, identify bugs, and allow the Assistant to fix them automatically.

“With new capabilities in planning, building, and testing, Assistant better utilizes agentic loops to evaluate game aspects, offer suggested solutions, and incorporate results into future planning loops, forming a self-correcting system that improves over time,” Roblox conveyed in a blog post.

Roblox also announced plans to enable multiple AI agents to collaborate in parallel, conduct long workflows in the cloud, and manage tasks like coding, testing, and crafting more realistic game characters. They aim to ensure that creators can seamlessly integrate tools like Claude, Cursor, Codex, and others with Roblox Studio.

You might also like