MCP server for interfacing with Godot game engine. Provides tools for launching the editor, running projects, and capturing debug output.
Godot MCP has 20 tools with basic schemas and descriptions, but significant quality gaps limit production readiness. All tools have descriptions (10-50 chars, mostly under the 50-200 char LLM-optimized range), and input schemas are present with type definitions. However, descriptions are generic and lack actionable context (e.g., 'Launch the Godot editor with an optional project path' does not explain WHEN to use this vs run_godot_project, or what happens on success). Parameters lack format constraints, enums, and dependency documentation. Output schemas are entirely undocumented, LLMs cannot infer what fields to expect from responses. Error handling is minimal; no recovery guidance or categorization. Security concerns: file operations (read_file, write_file, delete_file) accept arbitrary paths with only basic '../' validation, risking directory traversal. No permission gates on destructive operations (delete_node, delete_file). Tool composition is reasonable (separate create/read/modify/delete), but missing pagination, result limits, and chaining IDs in responses.
Add a node to an existing scene
Add a texture to a node in a scene
Attach a script to a node in a scene
Create a new scene in a Godot project
Create a new script file in a Godot project
Delete a file from a Godot project
Delete a node from a scene
Export a Godot project to a specified platform
Get the captured debug output from the running Godot process
Output schemas completely undocumented. No tool documents what fields are returned, forcing LLMs to guess response structure and plan downstream calls blindly.
Descriptions are generic and lack actionable context. E.g., 'Get information about a Godot project' does not explain what fields are returned, when to call it vs list_scenes, or what prerequisites exist. Descriptions should be 50-200 chars and answer WHAT, WHEN, and WHAT-NEXT.
Path validation is insufficient. read_file, write_file, delete_file accept arbitrary paths with only '../' check. No whitelist, no project boundary enforcement. Malicious agents can read/write/delete files outside the intended project directory.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |
Get information about a Godot project
Get the node structure of a scene
Launch the Godot editor with an optional project path
List files in a directory within a Godot project
List all scenes in a Godot project
Modify a property of a node in a scene
Read the contents of a file in a Godot project
Rename a node in a scene
Run a Godot project and capture its output
Stop the currently running Godot process
Write contents to a file in a Godot project
Destructive operations (delete_node, delete_file) lack permission gates, confirmation steps, or dry-run support. An agent can irreversibly delete project assets without safeguards.
No pagination or result limits on list tools. list_scenes and list_files can return unbounded results, exhausting context windows. No documented limit or next_cursor support.