DOCUMENT PROCESSING / Docling Project
Docling MCP
Convert documents into structured Docling data, manipulate cached documents and optionally connect document-generation or retrieval tools.
“Convert documents into structured Docling data, manipulate cached documents and optionally connect document-generation or retrieval tools.”
01 / THE REASONING
Why this made the selection.
- Convert documents into structured Docling data, manipulate cached documents and optionally connect document-generation or retrieval tools.
- The setup documentation makes this operating boundary explicit: Remote conversion sends documents to the configured service; local conversion requires docling-mcp[local] and its model dependencies. Generation/RAG groups may need additional model or database credentials.
02 / THE REVIEW RECORD
What we actually inspected.
Source review has boundaries.
A clear record is more useful than a “safe” badge.
Material inspected
- README.md
- LICENSE
- pyproject.toml
- docling_mcp/servers/mcp_server.py
- docling_mcp/docling_cache.py
Our findings
- Remote conversion sends documents to the configured service; local conversion requires docling-mcp[local] and its model dependencies. Generation/RAG groups may need additional model or database credentials.
- The default transport is Streamable HTTP, so stdio clients must request --transport stdio. Local fallback is off by default and requires local extras; the bounded in-memory document cache is not persistent storage. Model licences are separate.
Not established by this review
- No server installation, authentication, live tool execution or isolation test was performed.
- Package/container parity with the pinned source and transitive dependency security were not established.
The review applies to the material and revision named here. A newer upstream release can change its behavior.
03 / PUT IT TO WORK
Connect a server deliberately.
DOCUMENTED COMMAND
uvx --from docling-mcp docling-mcp-server --transport stdio Copying does not execute this command. It may retrieve a newer version than the reviewed source.
- For a Docling Serve backend, set DOCLING_MCP_CONVERSION_MODE=remote, DOCLING_MCP_SERVICE_URL and DOCLING_MCP_SERVICE_API_KEY, then register uvx --from docling-mcp docling-mcp-server --transport stdio.
- Verify the connected account, available tools and target resource before authorizing mutations.
Before you start
- Pinned package Python requirement: >=3.10.
- A compatible local process MCP client.
- Remote conversion sends documents to the configured service; local conversion requires docling-mcp[local] and its model dependencies. Generation/RAG groups may need additional model or database credentials.
04 / FOLLOW THE EVIDENCE
The source trail.
Our notes are separate from the original resource.
Check upstream before adopting a new version.
https://github.com/docling-project/docling-mcp/blob/5e1551a9b984f9a26349de6d23d02eccf5a69db2/README.md
Supports: description, setup, capabilities, review, limitations
https://github.com/docling-project/docling-mcp/blob/5e1551a9b984f9a26349de6d23d02eccf5a69db2/LICENSE
Supports: license
- Pinned package metadata ↗Checked
https://github.com/docling-project/docling-mcp/blob/5e1551a9b984f9a26349de6d23d02eccf5a69db2/pyproject.toml
Supports: install, language, review
https://github.com/docling-project/docling-mcp/blob/5e1551a9b984f9a26349de6d23d02eccf5a69db2/docling_mcp/servers/mcp_server.py
Supports: review, install, access, limitations
https://github.com/docling-project/docling-mcp/blob/5e1551a9b984f9a26349de6d23d02eccf5a69db2/docling_mcp/docling_cache.py
Supports: review, install, access, limitations