Doc Extract MCP
MCP server giving AI agents safe PDF/text reading, chunking and schema-checked JSON output
- Pricing
- Open source
- Platforms
- CLI, self-hosted, API
- Version
- 0.1.0
- Listing updated
- today
Similar MCP servers
About Doc Extract MCP
Doc Extract MCP is an open-source MCP server for files and documents. It is available on CLI, self-hosted and API.
The README says the server handles the deterministic parts of document extraction that an LLM should not be trusted with: reading files, parsing PDFs, splitting long documents into overlapping chunks, validating JSON against a schema and writing results. The model still performs the actual field extraction; this server keeps it working from real text and checks the output.
All file access is confined to a single root folder set by the DOC_EXTRACT_ROOT variable, and paths or glob patterns that escape it are refused, including through symlinks. PDF text is returned with page markers so page references can be trusted. Schema validation reports every error with a JSON Pointer path in one pass rather than stopping at the first.
It is aimed at developers building document-automation workflows, such as turning purchase orders, invoices or reports into structured data. The project is a portfolio release, has tests, and the README itself notes it has not been hardened beyond its path guard and has no production history.
Tools this server provides (6)
Taken from the project documentation; the running server may differ.
- list_documents
- List files under a directory inside the allowed root, with size and modified time.
- read_document
- Return a document's text, with PDF page markers and optional page selection.
- document_info
- Return metadata for a document without its full content.
- chunk_document
- Split a document into ordered overlapping chunks.
- validate_json
- Validate a JSON string against a JSON Schema and return every error.
- save_structured
- Write extracted data as JSON or CSV inside the allowed root.
What it does
- Confined file access
- All reads and writes are restricted to one configured root folder, with traversal and symlink escapes refused.
- PDF and text reading
- Extracts text from PDFs with page markers and optional page ranges, plus .txt, .md, .json and .csv files.
- Deterministic chunking
- Splits documents into ordered chunks with configurable size and overlap, with page hints for PDFs.
- Full schema validation
- Checks JSON against Draft 2020-12 JSON Schema and returns every error with its path.
- Guarded output writing
- Saves results as JSON or CSV inside the root and reports row and byte counts.
Good for
- Extract line items from text-based invoices into schema-checked JSON
- Batch-process a folder of PDF reports for an AI extraction pipeline
- Give a Claude Code or Claude Desktop session safe access to a documents folder
- Split long contracts into overlapping chunks before passing them to an LLM
Pricing
Open source
MIT licensed, self-hosted from the GitHub repository.
Doc Extract MCP FAQ
Does it perform OCR on scanned PDFs?
No. The README says it uses pypdf for text-based PDFs only, and scanned or image-only PDFs yield empty text.
Which file types are supported?
The README lists .pdf, .txt, .md, .csv and .json. It states that .docx and .xlsx are not supported.
Which clients can use it?
The README says it pairs with Claude Code and Claude Desktop and with any other MCP client.
Does the server extract the data itself?
No. The README says the LLM performs the extraction reasoning, while the server handles file access, parsing, chunking and validation.
Is Doc Extract MCP free?
Yes. Doc Extract MCP is open source, so you can use and self-host it for free.
What are the best alternatives to Doc Extract MCP?
The closest alternatives to Doc Extract MCP are Pdf MCP, ReadAnyDoc, Docparser MCP by usefulapi, MCP File Processor and DocuSense.
Where this comes from
- Description: written from the project's README and its MCP Registry entry.