What it does
The server provides three tools to MCP-compatible clients such as Cursor, Claude Desktop and OpenClaw. The extract tool takes a document and a JSON schema, given as a file path or built inline by the assistant from the conversation, and returns structured data. It supports OCR with LLM extraction by default, plus fast, trained, robust and robust-trained modes. The to_markdown tool converts PDFs, images, DOCX, XLSX, CSV, EML and other supported files to markdown. The doctor tool checks configuration, API keys, Python dependencies and system binaries.
An assistant would reach for it when it needs fields pulled from invoices, forms or other documents into a structured format. The to_markdown tool helps inspect a document's contents before a schema is chosen, and doctor is meant for troubleshooting a setup that is not working.
By default the server runs in service mode, which uses the ClicheFactory cloud and needs one API key, set as CLICHEFACTORY_API_KEY or saved with the configure command. A local mode is available for users who bring their own LLM key. It requires installing the local extras, system binaries such as tesseract and LibreOffice, and Python 3.12 or later. Batch operations, interactive configuration and file output from the CLI are not included in version 1.
Limitations and requirements
Service mode requires a ClicheFactory API key. Local mode requires an LLM key, the local extras (about 2 GB of parsing and OCR dependencies), system binaries (tesseract, LibreOffice) and Python 3.12 or later. Local mode quality depends on the user's setup. Batch operations, the configure command and the allow_partial parameter are not available in version 1. OCR engine selection uses the SDK default (RapidOCR) unless changed in the config file.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 14 Nov 2026 · its page on the source
ClicheFactory Document Intelligence
Extract structured JSON from PDFs, images, DOCX, XLSX, CSV, EML attachments, and DSPy pipelines.
v0.2.0As shown on registry.modelcontextprotocol.io when we last read it