Connects ClicheFactory's document extraction API to Claude and other MCP clients through three tools: extract, to_markdown, and doctor. You give it a file (PDF, DOCX, XLSX, images, EML) and a JSON schema, it returns structured data. Runs in two modes: local uses your own LLM key with OCR on your machine, service mode hits the ClicheFactory API and supports trained pipelines and robust verification. The to_markdown tool is handy for previewing documents before deciding what to extract. Reach for this when you need to pull invoice line items, form fields, or tabular data from documents without writing parsers.
claude mcp add --transport stdio clichefactory-clichefactory-mcp uvx clichefactory-mcp