Loading ToolBench
MCP server for social media analytics, video intelligence, and content creator tools. Provides access to SocialGPT's video analysis, creator comparison, and content performance auditing capabilities.
SocialGPT MCP provides 2 tools with generally clear names and well-structured input schemas. Both tools follow verb_noun naming conventions (list_videos, list_creator_videos) and have descriptive parameter enums. However, output schemas are not explicitly documented in the provided source, and error handling guidance is absent. The shared library (sgpt_lib.py) demonstrates sophisticated field-access and statistics implementation, suggesting the backend is mature, but the MCP tool interface itself lacks recovery guidance and explicit output documentation that LLMs need for chaining and error correction.
List videos from a specific creator/account on a given platform, with optional AI-powered analysis
List videos from the authenticated user's account, optionally including AI-powered analysis themes and metrics
Output schemas not documented. Tool descriptions state what the tools return (e.g., 'videos from the authenticated user's account') but do not formally document the response structure, field types, or pagination behavior. LLMs need explicit schema to understand what fields are available for downstream planning and extraction.
No error handling or recovery guidance. Neither tool description mentions failure modes, invalid inputs, or how to recover from errors. E.g., what happens if a creator username doesn't exist? What if authentication fails? Agents need actionable error messages.
Parameter descriptions lack format/constraint details. The 'limit' parameter states 'default: 15, max: 50' but does not explain what happens if the agent passes an invalid value (e.g., limit=999 or limit=-1). Add explicit validation rules: 'Must be an integer between 1 and 50; defaults to 15.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | 2026-07-28+ | v2 |
Tool descriptions are generic and lack 'when to use' context. Both tool descriptions state what they do but not when an agent should call list_videos vs list_creator_videos. E.g., 'Call list_videos to fetch your own account videos; use list_creator_videos to analyze a competitor's or public creator's channel.'
Missing pagination details in tool descriptions. Both tools accept a 'limit' parameter but do not mention how to retrieve the next page of results. Does the response include a 'next_cursor', 'offset', or 'total_count'? Agents cannot navigate large result sets without explicit pagination guidance.
Tool composition is asymmetric. list_videos requires authentication (described as 'authenticated user's account') but does not explicitly state this prerequisite. list_creator_videos accepts a 'platform' parameter but list_videos does not, it appears to assume the user's default platform. Clarify whether list_videos is platform-specific.