Loading ToolBench
AI-powered screen recording editor that applies auto-zoom and editing to demo videos
This MCP server has ONE tool with critical definition gaps. The tool name is acceptable (verb_noun format), but the schema is severely underdocumented, parameter descriptions are missing or trivial, and there is no documented output schema structure. The entire server appears to be a mock/stub implementation with hardcoded return values, not a real integration. No MCP transport layer is visible in the source code, only a bare Python client class. The server cannot be evaluated as an MCP tool at all without evidence of MCP protocol implementation (no mcp.Server, no tool_call handlers, no JSON-RPC transport).
Edits a raw screen recording video with auto-zoom keyframes and returns the processed video URL and keyframe count
Tool schema lacks parameter descriptions. 'raw_screen_video_path' and 'output_aspect_ratio' have no descriptions explaining what format/constraints are expected or how they affect behavior.
No documented output schema. The tool returns {'edited_video_url': str, 'auto_zoom_keyframes_count': int}, but LLMs have no explicit definition of these fields' types or semantics. What does 'auto_zoom_keyframes_count' represent? Is the URL always valid/accessible?
Parameter 'output_aspect_ratio' has a default ('16:9') but no enum or validation constraint. LLMs may pass invalid values (e.g. '4:3', '21:9', 'square') with no guidance. The description does not constrain valid options.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 30 | 2026-07-28+ | v2 |
Tool description is generic and lacks context. 'Edits a raw screen recording video with auto-zoom keyframes and returns the processed video URL and keyframe count' does not explain WHEN to use this tool, what 'auto-zoom' means, or what prerequisites exist (e.g. must raw_screen_video_path be a local file or remote URL?).
No MCP protocol implementation visible in source. The repository contains only a bare Python client class (ScreenRecordingDemoAiEditorClient) and example usage. There is no mcp.Server, no tool registration, no JSON-RPC transport, no HTTP/SSE server, and no evidence of MCP protocol support. This appears to be a demo stub, not a functional MCP server.
Tool performs file I/O (reads video, writes processed video) but has no error handling for file not found, permission denied, disk full, or invalid video format. Error responses would be bare exceptions, not actionable guidance for LLM recovery.
The implementation is a mock stub returning hardcoded values (hash-based fake URL, static keyframe count of 14). No actual video processing occurs. This is not suitable for production use and misleads about tool capabilities.