Playforge runs Claude-powered agents inside Roblox Studio. They play every build, hunt bugs, check mobile layouts and flag where new players get lost, then hand you a report before you publish.
Most Roblox studios ship every week and test by clicking around in Studio or asking friends on Discord.
Discovery depends on early-session retention. One broken update can sink a game in the algorithm.
Outsourced testers are slow, costly and can't run on every commit.
Link your Rojo project or open place. Playforge starts a test session through the Roblox Studio MCP.
A new mobile player, a grinder, an exploit hunter. Each agent plays, inspects game state and takes screenshots.
Agents check real state: currency, inventory, positions, remote calls. Every bug comes with repro steps.
Bugs, onboarding friction, broken mobile UI and exploitable remotes. Runs in CI on every push.
Illustrative example of the report format.
Playtesting is a multi-step loop: play, form a hypothesis, check it, repeat. That is exactly where Claude's tool use shines.
Agents read the Luau behind a bug and propose the patch, not just the symptom.
Vision checks UI layout across phone, tablet and desktop sizes.
Playforge grows out of tooling we already use to build and test our own Roblox games: in-house frameworks and agent-driven QA through Roblox Studio. We're opening early access to a small group of studios.