Skip to main content
Glama

Docling MCP: making docling agentic

CI PyPI version PyPI - Python Version uv Ruff Pydantic v2 prek License MIT PyPI Downloads LF AI & Data MCP Registry docling-mcp MCP server MCP Toplist

A document processing service using the Docling-MCP library and MCP (Model Context Protocol) for tool integration.

Overview

Docling MCP is a service that provides tools for document conversion, processing and generation. It uses the Docling library to convert PDF documents into structured formats and provides a caching mechanism to improve performance. The service exposes functionality through a set of tools that can be called by client applications.

Related MCP server: PDF MCP Server

Compatibility

docling-mcp

MCP Python SDK

>=3.0.0

mcp>=2.0.0

>=2.0.0,<3.0.0

mcp>=1.9.4,<2.0.0

If your MCP client application has not yet migrated to MCP SDK v2, pin:

pip install "docling-mcp<3.0.0"

See MIGRATION.md for the full migration guide.

Installation Options

For users with access to Docling Serve API:

Getting Docling Serve: Visit docling-serve for installation guides. You can deploy it from published container images or look for managed Docling SaaS offerings.

pip install docling-mcp

Then configure your environment:

export DOCLING_MCP_SERVICE_URL=https://your-docling-service.example.com
export DOCLING_MCP_SERVICE_API_KEY=your-api-key-here
export DOCLING_MCP_CONVERSION_MODE=remote

Local Mode (Full Features)

For users who need local conversion or don't have Docling Serve access:

pip install docling-mcp[local]

Then configure your environment:

export DOCLING_MCP_CONVERSION_MODE=local

Hybrid Mode (Best of Both)

Install with local support and enable automatic fallback:

pip install docling-mcp[local]

Configure for remote with fallback:

export DOCLING_MCP_SERVICE_URL=https://your-docling-service.example.com
export DOCLING_MCP_CONVERSION_MODE=remote
export DOCLING_MCP_FALLBACK_TO_LOCAL=true

Features

  • Conversion tools:

  • Generation tools:

    • Document generation in DoclingDocument, which can be exported to multiple formats

  • Local document caching for improved performance

  • Support for local files and URLs as document sources

  • Memory management for handling large documents

  • Logging system for debugging and monitoring

  • RAG applications with Milvus upload and retrieval

Configuration

All settings use the DOCLING_MCP_ prefix and can be supplied as environment variables, in a .env file in the working directory, or via the env block of your MCP client config. Copy .env.example as a starting point.

Conversion mode

Variable

Default

Description

DOCLING_MCP_CONVERSION_MODE

remote

remote or local

Remote service (required when DOCLING_MCP_CONVERSION_MODE=remote)

Variable

Default

Description

DOCLING_MCP_SERVICE_URL

—

URL of the Docling Serve instance

DOCLING_MCP_SERVICE_API_KEY

—

API key for the service

DOCLING_MCP_SERVICE_TIMEOUT

300.0

Timeout in seconds for a whole conversion job

DOCLING_MCP_SERVICE_MAX_RETRIES

3

Max retry attempts

DOCLING_MCP_FALLBACK_TO_LOCAL

false

Fall back to local if service is unreachable (requires docling-mcp[local])

Conversion pipeline (applies to both modes)

Variable

Default

Description

DOCLING_MCP_KEEP_IMAGES

false

Retain page images in output

DOCLING_MCP_IMAGES_SCALE

1.0

Image scale factor (increase to avoid tensor padding errors)

DOCLING_MCP_DO_OCR

true

Run OCR pipeline

DOCLING_MCP_OCR_ENGINE

auto

OCR engine by its docling engine kind: auto, easyocr, tesseract (system binary), tesserocr (Python bindings), ocrmac, rapidocr. The engine must be installed on the conversion backend. In remote mode it is sent as the Docling Serve ocr_preset, so server-defined preset names also work there

DOCLING_MCP_OCR_LANG

—

Comma-separated OCR languages, passed to the engine unchanged. iso:en,iso:de works with any engine (docling ≥ 2.127, recommended with auto); bare codes are engine-specific (eng for Tesseract, en for EasyOCR or Docling Serve). Unset means the engine's defaults

DOCLING_MCP_DO_TABLE_STRUCTURE

true

Detect table structure

DOCLING_MCP_DO_PICTURE_DESCRIPTION

false

Run a vision-language model to generate textual descriptions for figures and pictures

DOCLING_MCP_PICTURE_DESCRIPTION_PRESET

—

Preset ID for the picture-description VLM (e.g. smolvlm, granite_vision). When unset the backend default is used. Only applies when DOCLING_MCP_DO_PICTURE_DESCRIPTION=true

TIP

EnablingDOCLING_MCP_DO_PICTURE_DESCRIPTION=true adds a VLM inference step for every figure in the document. On resource-constrained hardware (e.g. a laptop running the model on CPU) this can significantly increase conversion time.

  • Remote mode — increase DOCLING_MCP_SERVICE_TIMEOUT (in seconds) to give the remote service enough time to complete. The default is 300.0 (5 minutes).

  • Local mode / MCP client timeout — the MCP client itself may also enforce a request timeout. For example, in LM Studio set the timeout field (in milliseconds) in your mcp.json:

{
  "mcpServers": {
    "docling": {
      "url": "http://localhost:8000/mcp",
      "timeout": 300000
    }
  }
}

Markdown export

Variable

Default

Description

DOCLING_MCP_IMAGE_EXPORT_MODE

placeholder

How images are rendered in Markdown output: placeholder (emits <!-- image -->), embedded (base64 data-URI), referenced (file path / URL)

Cache

Variable

Default

Description

DOCLING_MCP_CACHE_MAX_DOCUMENTS

10

Maximum number of documents to keep in the in-memory cache. When the limit is reached the least-recently-used document is evicted.

LlamaIndex RAG (--tools llama-index-rag)

Variable

Default

Description

DOCLING_MCP_LI_API_BASE

http://127.0.0.1:1234/v1

OpenAI-compatible LLM endpoint

DOCLING_MCP_LI_API_KEY

none

API key for the LLM endpoint

DOCLING_MCP_LI_MODEL_ID

ibm/granite-3.2-8b

LLM model identifier

DOCLING_MCP_LI_EMBEDDING_MODEL

BAAI/bge-base-en-v1.5

HuggingFace embedding model

LlamaStack (--tools llama-stack-rag / --tools llama-stack-ie)

Variable

Default

Description

DOCLING_MCP_LLS_URL

http://localhost:8321

LlamaStack server URL

DOCLING_MCP_LLS_VDB_EMBEDDING

all-MiniLM-L6-v2

Embedding model for vector DB

DOCLING_MCP_LLS_EXTRACTION_MODEL

openai/gpt-oss-20b

Model used for structured extraction

Setting variables in an MCP client config

{
  "mcpServers": {
    "docling": {
      "command": "uvx",
      "args": [
        "--from=docling-mcp",
        "docling-mcp-server"
      ],
      "env": {
        "DOCLING_MCP_CONVERSION_MODE": "remote",
        "DOCLING_MCP_SERVICE_URL": "https://your-docling-service.example.com",
        "DOCLING_MCP_SERVICE_API_KEY": "your-api-key-here"
      }
    }
  }
}

Getting started

The easiest way to install Docling MCP and connect it to your client is by launching it via uvx.

Depending on the transfer protocol required, specify the argument --transport, for example

  • stdio used e.g. in Claude for Desktop and LM Studio

    uvx --from docling-mcp docling-mcp-server --transport stdio
  • sse used e.g. in Llama Stack

    uvx --from docling-mcp docling-mcp-server --transport sse
  • streamable-http used e.g. in containers setup

    uvx --from docling-mcp docling-mcp-server --transport streamable-http

More options are available, e.g. the selection of which toolgroup to launch. Use the --help argument to inspect all the CLI options.

For developing the MCP tools further, please refer to the Developing section of CONTRIBUTING.md for instructions.

Integration with MCP clients

One of the easiest ways to experiment with the tools provided by Docling MCP is to leverage an AI desktop client with MCP support. Most of these clients use a common config interface. Adding Docling MCP in your favorite client is usually as simple as adding the following entry in the configuration file.

{
  "mcpServers": {
    "docling": {
      "command": "uvx",
      "args": [
        "--from=docling-mcp",
        "docling-mcp-server"
      ]
    }
  }
} 

When using Claude for Desktop, simply edit the config file claude_desktop_config.json with the snippet above or the example provided here.

In LM Studio, edit the mcp.json file with the appropriate section or simply click on the button below for a direct install.

Add MCP Server docling to LM Studio

Other integrations are described in the integrations page.

Examples

Converting documents

Example of prompt for converting PDF documents:

Convert the PDF document at <provide file-path> into DoclingDocument and return its document-key.

Generating documents

Example of prompt for generating new documents:

I want you to write a Docling document. To do this, you will create a document first by invoking `create_new_docling_document`. Next you can add a title (by invoking `add_title_to_docling_document`) and then iteratively add new section-headings and paragraphs. If you want to insert lists (or nested lists), you will first open a list (by invoking `open_list_in_docling_document`), next add the list_items (by invoking `add_list_items_to_list_in_docling_document`). After adding list-items, you must close the list (by invoking `close_list_in_docling_document`). Nested lists can be created in the same way, by opening and closing additional lists.

During the writing process, you can check what has been written already by calling the `export_docling_document_to_markdown` tool, which will return the currently written document. At the end of the writing, you must save the document and return me the filepath of the saved document.

The document should investigate the impact of tokenizers on the quality of LLMs.

Contributing

We welcome external contributions. See CONTRIBUTING.md for details on how to get started.

License

The Docling MCP codebase is under MIT license. For individual model usage, please refer to the model licenses found in the original packages.

LF AI & Data

Docling and Docling MCP is hosted as a project in the LF AI & Data Foundation.

IBM ❤️ Open Source AI: The project was started by the AI for knowledge team at IBM Research Zurich.

Available Tools

21 tools
add_list_items_to_list_in_docling_documentAdd items to list in Docling documentA

Add list items to an open list in an existing document in the local document cache.

This tool inserts new list items with the specified text and marker into an open list within a document. It requires that the document exists and that there is at least one item in the document's stack cache.

Inputs: list_items (required), document_key (required). Returns: document_key

add_paragraph_to_docling_documentAdd paragraph to Docling documentB

Add a paragraph of text to an existing document in the local document cache.

This tool inserts a new paragraph under the specified section header and level into a document that has already been processed and stored in the cache.

Inputs: paragraph (required), document_key (required). Returns: document_key

add_section_heading_to_docling_documentAdd section heading to Docling documentA

Add a section heading to an existing document in the local document cache.

This tool inserts a section heading with the specified heading text and level into a document that has already been processed and stored in the local cache. Section levels typically represent heading hierarchy (e.g., 1 for H1, 2 for H2).

Inputs: document_key (required), section_level (required), section_heading (required). Returns: document_key

add_table_in_html_format_to_docling_documentAdd HTML table to Docling documentA

Add an HTML-formatted table to an existing document in the local document cache.

This tool parses the provided HTML table string, converts it to a structured table representation, and adds it to the specified document. It also supports optional captions and footnotes for the table.

Inputs: html_table (required), document_key (required), table_captions, table_footnotes. Returns: document_key

add_title_to_docling_documentAdd or update title to Docling documentA

Add or update the title of a document in the local document cache.

This tool modifies an existing document that has already been processed and stored in the local cache. It requires that the document already exists in the cache before a title can be added.

Inputs: title (required), document_key (required). Returns: document_key

check_document_in_local_cacheCheck if Docling document is in cacheA
Read-only

Check whether a Docling document is already converted and present in the local cache.

Inputs: document_key (required). Returns: in_cache

close_list_in_docling_documentClose list in Docling documentA

Closes a list group in an existing document in the local document cache.

This tool closes a previously opened list structure within a document. It requires that the document exists and that there is more than one item in the document's stack cache.

Inputs: document_key (required). Returns: document_key

convert_directory_files_into_docling_documentConvert files from directory into Docling documentA

Convert all files from a local directory path and store them in local cache.

This tool takes a local directory path, converts every file in the directory using the configured converter (remote API or local) and stores the resulting Docling documents in a local cache. It returns a list of conversion outputs, where each output consists of a boolean set to False along with a document's unique cache key. If a document was already in the local cache, the conversion is skipped and the output boolean is set to True.

Inputs: source (required). Returns: result

convert_document_into_docling_documentConvert document into Docling documentA

Convert a document from a URL or local path into a Docling document.

Use this tool when you have an existing file (PDF, DOCX, HTML, image, etc.) that you want to load and parse.

This tool takes a document's URL or local file path, converts it using the configured converter (remote API or local), and stores the resulting Docling document in a local cache. It returns an output with a boolean set to False along with the document's unique cache key. If the document was already in the local cache, the conversion is skipped and the output boolean is set to True.

Inputs: source (required). Returns: from_cache, document_key

create_new_docling_documentCreate new Docling documentA

Create a new, empty Docling document ready for incremental authoring.

Use this tool when you want to build a document from scratch by subsequently calling add_title_to_docling_document, add_section_heading_to_docling_document, add_paragraph_to_docling_document, open_list_in_docling_document, and related tools. To convert an already-existing file (PDF, DOCX, HTML, etc.) instead, use convert_document_into_docling_document.

The document is stored in the local cache under a unique key and the prompt text is recorded as a furniture item for reference.

Inputs: prompt (required). Returns: prompt, document_key

delete_document_items_at_anchorsDelete Docling document items at anchorsA
Destructive

Delete multiple document items identified by their anchors.

This tool removes specified items from a Docling document that exists in the local document cache, based on their anchor references. It requires that the document already exists in the cache before performing the deletion.

Inputs: document_key (required), document_anchors (required). Returns: document_key

export_docling_document_to_markdownExport Docling document to markdown formatA
Read-only

Export a document from the local document cache to markdown format.

This tool converts a Docling document that exists in the local cache into a markdown formatted string, which can be used for display or further processing. Set page_no to export a single page instead of the whole document.

Inputs: document_key (required), page_no, max_size. Returns: markdown, document_key

get_overview_of_document_anchorsGet overview of Docling document anchorsA
Read-only

Retrieve a structured overview of a document from the local document cache.

This tool returns a text representation of the Docling document's structure, showing the hierarchy and types of elements within the document. Each line in the output includes the document anchor reference and item label.

Inputs: document_key (required). Returns: structure

get_page_thumbnailGet page thumbnail of Docling documentB
Read-only

Generate a thumbnail image for the requested page of a document in the local cache.

Inputs: document_key (required), size, page_no.

get_text_of_document_item_at_anchorGet text of Docling document item at anchorA
Read-only

Retrieve the text content of a specific document item identified by its anchor.

This tool extracts the text from a Docling document item at the specified anchor location within a document that exists in the local document cache.

Inputs: document_key (required), document_anchor (required). Returns: text

list_cached_documentsList cached Docling documentsA
Read-only

Return the keys of all Docling documents currently held in the local cache.

Use this tool to discover which documents are available before calling any tool that requires a document_key. An empty list means the cache is empty and a document must first be loaded with convert_document_into_docling_document or created with create_new_docling_document.

Inputs: no named parameters. Returns: document_keys

open_list_in_docling_documentOpen list in Docling documentA

Open a new list group in an existing document in the local document cache.

This tool begins a list context on the document's internal stack. Once a list is open, only add_list_items_to_list_in_docling_document may be called to populate it. Attempting to add a title, section heading, paragraph, or another list before calling close_list_in_docling_document will raise an error. Nested lists are supported: call open_list_in_docling_document again after adding items to start an inner list, then close each list with close_list_in_docling_document in reverse order.

Inputs: document_key (required). Returns: document_key

remove_document_from_local_cacheRemove document from local cacheA
Destructive

Remove a document from the local cache and release its memory.

Call this tool when a client is finished with a document and wants to release the memory it occupies. After a successful call the key is no longer present in the cache.

Inputs: document_key (required). Returns: dropped

save_docling_documentSave Docling documentA

Save a document from the local document cache to disk in both markdown and JSON formats.

This tool takes a document that exists in the local cache and saves it to the specified cache directory with filenames based on the document key. Both markdown and JSON versions of the document are saved. When page_no is set, only that page is saved in markdown, to a page-specific file, while the JSON file still contains the full document.

Inputs: document_key (required), page_no. Returns: md_file, json_file

search_for_text_in_document_anchorsSearch for text in Docling document anchorsA
Read-only

Search for specific text and keywords within a document's anchors.

This tool takes a string of text to search for and returns a string of all document anchors that contain the exact text. The search is case-insensitive. If the exact text is not found, the tool will search for individual keywords within the text, splitting it on non-alphanumeric characters. If keywords are found, they are listed alongside their number of occurrences in parentheses.

Inputs: text (required), document_key (required). Returns: result

update_text_of_document_item_at_anchorUpdate text of Docling document item at anchorA
Destructive

Update the text content of a specific document item identified by its anchor.

This tool modifies the text of an existing document item at the specified anchor location within a document that exists in the local document cache. It requires that the document already exists in the cache before a modification can be made.

Inputs: document_key (required), updated_text (required), document_anchor (required). Returns: document_key

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 21 tools

    Disambiguation4/5

    Most tools have clearly distinct resource+action targets, and the descriptions actively disambiguate close pairs (e.g. create_new_docling_document explicitly says to use convert_document_into_docling_document for existing files). Minor overlap remains between check_document_in_local_cache and list_cached_documents, and between export_docling_document_to_markdown and save_docling_document, but the distinctions are documented.

    Naming Consistency4/5

    Names are uniformly snake_case and verb-first (check_, convert_, add_, get_, delete_, etc.), which is readable and predictable. However, the object naming is inconsistent: some tools append the full '_to_docling_document' suffix while others use plain 'document', 'cached_documents', or 'page_thumbnail'.

    Tool Count4/5

    At 21 tools it is on the heavy side, but the surface splits into coherent clusters (cache, conversion, authoring, inspection, export), and nearly every tool maps to a distinct operation. The list authoring stack (open/close/add_items) inflates the count slightly but is justified.

    Completeness4/5

    The lifecycle is well covered: conversion, cache listing/checking/removal, document creation, structural authoring (title, headings, paragraphs, lists, tables), anchor-based inspection/search/update/delete, and export/save/thumbnail. Gaps are minor, e.g. no whole-document text extraction or format export beyond markdown/JSON, and no item reordering or bulk text replacement.

    Maintenance

    ActivityActive
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      D
      maintenance
      Enables processing and analysis of large PDF files through text extraction, search functionality, and intelligent chunking strategies. Provides comprehensive PDF operations including metadata retrieval, page-range text extraction, and content search with contextual results.
      -
    • A
      license
      A
      quality
      D
      maintenance
      Enables PDF processing and analysis including text extraction, metadata retrieval, search, page manipulation, splitting/merging, conversion to images, and form handling.
      10
      MIT