Roblox Studio MCP

When to Load

Load for Studio MCP work: scripts, scenes, assets, debugging, or playtesting. Skip for standalone code generation.

Quick Reference

Route by capability first. Two Studio MCP bridges exist: the official (built into Studio, closed-source) and chrrxs's (MIT). Route on tools/list (getconnectedinstances/eval/multiplayer → chrrxs; listrobloxstudios + per-call studio_id → official) BEFORE bridge-specific bootstrap.

Bootstrap before mutation

1. Identify the target via listrobloxstudios (official) or getconnectedinstances (chrrxs).
2. Pass the target's studioid (official) or instanceid (chrrxs) on every tool call.
3. getstudiostate and confirm Edit/Client/Server availability.
4. Inspect the target tree and scripts before changing them.

Execution contract

discover → select Studio/context → inspect → mutate in bounded batches
→ read back → start play → evaluate live → evidence → clean up

Edit-time injection (multiedit, Edit-context executeluau) writes scripts; it never targets a running playtest. Live evaluation targets a running VM: start play, wait for the intended Client/Server VM, then evaluate there. A stopped VM is never a live-evaluation target. Pass datamodel_type only where the tool requires it; do not guess from a previous session.

Capabilities

Inspect (searchgametree, inspectinstance, scriptread, scriptgrep), edit/execute (multiedit, executeluau), assets (searchasset/insertasset, generate* with waitjobfinished, storeimage/uploadimage), play/evidence (startstopplay, getconsoleoutput, screen_capture, input simulation, subagent).

Reliability rules

• execute_luau is stateless: re-acquire references every call.
• Read before write; read back after every script, asset, or geometry mutation.
• Choose reuse vs generation vs native fallback before asset work.
• Call waitjobfinished before edits that depend on a generation job.
• Command-size limits are bridge-specific. Split large scripts into separate bounded multiedit/executeluau writes (or supported multi-edit operations), then read back the full payload to verify.
• If MCP is absent, provide offline Luau and state what was not verified.

Full tool mappings, live schemas, asset workflows, and recovery rules: [references/full.md](references/full.md)