Roblox Studio MCP

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 (get_connected_instances/eval_*/multiplayer_* → chrrxs; list_roblox_studios + per-call studio_id → official) BEFORE bridge-specific bootstrap.

Bootstrap before mutation

1. Identify the target via list_roblox_studios (official) or get_connected_instances (chrrxs).
2. Pass the target's studio_id (official) or instance_id (chrrxs) on every tool call.
3. get_studio_state 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 (multi_edit, Edit-context execute_luau) 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 (search_game_tree, inspect_instance, script_read, script_grep), edit/execute (multi_edit, execute_luau), assets (search_asset/insert_asset, generate_* with wait_job_finished, store_image/upload_image), play/evidence (start_stop_play, get_console_output, 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 wait_job_finished before edits that depend on a generation job.
• Command-size limits are bridge-specific. Split large scripts into separate bounded multi_edit/execute_luau 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