Loading ToolBench
OCR.space MCP — wraps the OCR.space API (ocr.space) for image/PDF → text OCR.
Single tool with clear naming (extract_text) and good description (156 chars, within 10-1024 baseline). Input schema is complete with all parameters typed and described. However, output schema is undocumented, LLMs cannot predict the response structure (text, pages, exit_code fields). Error handling returns structured objects but lacks recovery guidance. No parameter constraints (enums) despite engine accepting only 1-3. Missing documentation of API key fallback behavior and rate limits.
Extract text from an image or PDF via OCR — pass an image/PDF URL, get the recognized text. Useful for reading screenshots, scanned documents, receipts, signs. Engine 2 is best for most cases.
Output schema undocumented. Tool returns {text, pages, exit_code} but LLMs cannot predict this structure. Breaks downstream tool chaining and forces LLMs to guess field names.
Engine parameter accepts 1, 2, or 3 but defined as free-form number type. Should be enum to prevent invalid values and self-document valid options.
Error handling returns structured errors but lacks recovery guidance. 'api_key_required' error does not explain how to provide a key or what the fallback is. 'ocr_failed' does not suggest retry or alternative actions.
Language parameter accepts free-form strings (e.g. 'eng', 'fre', 'ger', 'spa') but no enum constraint. LLMs may hallucinate invalid codes like 'english' or 'french'. Should enumerate supported languages.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 66 | 2026-07-28+ | v2 |
Transport mechanism unknown. Code exports a custom McpToolExport interface but does not show how it integrates with MCP protocol (STDIO, HTTP, SSE). Cannot verify protocol compliance.