Loading ToolBench
MCP server providing Pine Script v6 documentation for AI assistants
Pine Script MCP server demonstrates solid tool definition quality with all 7 tools having clear, action-verb names and documented input schemas. Descriptions are concise and functional (80-180 chars average), meeting LLM readability standards. All tools properly declare READ_ONLY risk. However, output schemas are not formally documented in the visible code, parameter descriptions are minimal for pagination parameters (limit/offset lack range constraints), and error handling guidance is absent. The server follows good naming conventions (list_, get_, search_, find_) and provides pagination support, but lacks structured error recovery patterns and output schema documentation that would elevate it to 80+. Tool definitions are explicit and visible in source; no inference required.
Find Pine Script v6 functions and namespaces by name
Read full content of a Pine Script documentation file
Get Prometheus metrics for monitoring tool usage and performance
Read a specific section from a Pine Script documentation file
List all available Pine Script documentation files with descriptions
Search Pine Script documentation by keyword or regex pattern
Semantic search through Pine Script documentation using AI embeddings
Output schemas not formally documented in code. While tools return structured data (documented implicitly in docstrings), there is no explicit JSON Schema definition visible for tool responses. LLMs need explicit return type definitions to plan downstream operations and extract fields accurately.
Pagination parameters (limit, offset) lack range constraints in descriptions. Best practice: state 'Maximum number of results to return (1 - 100, default 20)' instead of just 'Maximum number of results to return'. Unbounded numeric parameters allow LLMs to pass absurd values.
Error handling and recovery guidance absent. Tools do not document what errors are possible (e.g., 'file not found', 'invalid regex'), what each means, or what the LLM should do next. Pattern: error responses should guide recovery ('Try search_docs() with a simpler query').
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | B | 79 | 2025-06-18+ | v2 |
search_docs_semantic tool uses AI embeddings but does not document the embedding model, dimension, or similarity threshold. LLMs cannot reason about relevance quality or failures without this metadata.