Roblox Studio is Going Agentic

Today, 44% of the top 1,000 creators on Roblox use Roblox Assistant or third-party AI tools via MCP to plan, build, and test their games.¹ We’re leveling up Roblox Studio and Assistant with new agentic features so that creators can use AI to accelerate every step of the plan, build, test loop. We’re also surfacing features to third-party tools via Studio’s built-in MCP server. Creators now have more flexibility to augment their workflows with AI tools that can help them ship ambitious games faster.

Plan: Improved Planning Mode  

Game development is a multistep, multidiscipline endeavor that requires continual exploration and iteration. AI tools that intake a prompt and output a solution in one step can feel like shooting in the dark and often fail to truly capture a creator’s original intent. We’re introducing an improved Planning Mode to turn Assistant into a multistep, collaborative development partner that works with creators to analyze the game’s code and data model, ask clarifying questions, and turn a complex prompt into a highly detailed, reviewable, editable action plan.

Creators can add context and tweak the plan to make sure it reflects their intent before any changes are made. The plan serves as a mini game design document that agents can use to execute tasks in parallel and check their work against the original vision. Assistant creates a structured manifest of tasks with context that it can reference for the remainder of the session. Soon after launch, we’ll add the ability to automatically store this context so it can be referenced across sessions.

Build: New Mesh and Procedural Generation Tools

With a plan in place, it’s time to build out the world. New Mesh and Procedural Model Generation features accelerate the building process and help creators turn their plans into detailed, interactive worlds. Mesh Generation allows creators to quickly add textured meshes to the game world.

We’ll also soon introduce Procedural Models controlled by code. Creators will be able to generate Procedural Models with customizable attributes using text and image prompts, or they can build them from scratch. Attributes like the number of shelves in a bookcase, the number of chairs around a table, and more can be adjusted dynamically, creating smarter, editable building blocks that can be refined and reused for other purposes. 

Test: New Playtesting Agent Beta

Testing and iteration are key to creating a robust game that keeps players coming back. Many creators have built sophisticated testing workflows and envisioned even more sophisticated workflows enabled by AI:

Assistant’s new playtesting agent beta can help test the game against the original plan, analyzing the code and data model, reading logs, and using the player character as an automated QA tester to verify behavior. 

With the new capabilities across planning, building, and testing, Assistant is better at using agentic loops to test different aspects of the game, surface suggested solutions, and then incorporate the results into future planning loops, creating a self-correcting system that becomes more accurate over time. 

More Workflows, More Flexibility 

AI is accelerating workflows for planning, building, and testing games, and we’re committed to improving our tools so creators can close the gap between creative vision and execution. 

We’re working on enabling other common workflows, allowing agents to run in parallel, building long-form cloud agent workflows for complex tasks, developing more intelligent NPCs that can simulate more types of player behavior, visualizing AI workflows with a node graph, and supporting more tools and input methods. We’re also making sure creators can seamlessly use Claude, Cursor, Codex, and other third-party tools with Studio by surfacing all the context about a project via unprivileged APIs and Studio’s built-in MCP server.