Connects Claude to the space-ocr API for structured OCR that anchors LLM extractions to real Google Vision API character bounding boxes instead of hallucinated coordinates. Exposes two tools: ocr_extract for pulling structured fields from document images (receipts, invoices, passports, or custom schemas) and list_templates for browsing the nine built-in document types. Accepts URLs, local paths, or base64 images. Reach for this when your agent needs to auto-fill forms, build verification UIs, or reconcile accounting data and the pixel-level position of each character actually matters. Requires a space-ocr API key and runs via npx with flat ¥10 per-call pricing.
claude mcp add --transport stdio io.github.hwang-yh-cto-space-ocr -- npx -y space-ocr-mcp