Open-source, full-featured MCP server for AI-driven Godot game development, testing, and debugging
Single tool 'godot_exec' has a descriptive name and clear purpose, but critical gaps undermine usability. The tool description (127 chars) is adequate but the input schema is severely underdeveloped: 'action' and 'params' are both generic strings/objects with minimal type information. No output schema is documented. The description mentions three actions (eval, call, assert) but provides no parameter details for each, forcing LLMs to guess at required/optional fields. No error handling guidance. The tool is write-gated (good security), but lacks the structured input/output patterns needed for reliable agent composition.
Execute GDScript in running game (gated). Actions: eval(code,await?),call(node_path,method,args?),assert(condition,description?,await?). Disabled if --no-eval.
Input schema uses generic 'object' type for params with no field definitions. LLMs cannot determine which fields are required, their types, or valid values for each action (eval/call/assert).
No output schema documented. Callers cannot know what fields to expect in the response, breaking downstream tool composition and forcing LLMs to parse unstructured output.
Tool description mentions three actions (eval, call, assert) but provides zero parameter guidance for each. LLMs must infer that eval needs 'code' and 'await', call needs 'node_path' and 'method', etc.
No error handling guidance. If eval is disabled (--no-eval), the tool returns a permission error, but no recovery path is documented. LLMs cannot determine whether to retry, ask the user, or abandon the task.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
STDIO transport only. Server is not remotely accessible and cannot be used by hosted MCP clients. This is a hard architectural limitation.