Running GroupDocs MCP servers on-premise: architecture and security model Leave feedback

Run document search for AI agents fully on-premise: the GroupDocs.Search MCP server uses local stdio transport with no external endpoints, no inbound ports, and no telemetry, and keeps its index on your disk. This page is the one to send your security reviewer.

Note

The commands and config snippets on this page are for the .NET build of the server, which ships as a Docker image only. Installation and client setup: MCP server for .NET. Other platforms will expose the same tools with their own launch command; everything else on this page applies unchanged.

+--------------+        +----------------------+        +--------------------+
|  AI client   | stdio  |  Search MCP server   | reads  | documents          |
| Claude, VS   |<------>|  Docker container,   |<-------| /data: contracts/, |
| Code, Cursor |  JSON- |  GroupDocs.Search    |        | mail/, reports/    |
|              |  RPC   |  engine in-process   | reads/ +--------------------+
+--------------+        |                      | writes +--------------------+
                        |                      |<------>| index store        |
                        +----------------------+        | /data/.groupdocs-  |
                                                        | search-index/<name>|
  nothing listens on a port; no network at runtime      +--------------------+

• Transport: stdio to a child process — your AI client runs docker run --rm -i …. The server never listens on a network socket.
• Data path: agent → local container → mounted folders. Documents are read from the volume you mounted; nothing is transmitted anywhere.
• State: the index — extracted text and term data — is written to /data/.groupdocs-search-index or to the folder in GROUPDOCS_MCP_INDEX_PATH. Treat it with the same classification as the documents: it contains their text.
• Network use: only to pull the image (ghcr.io / docker.io). At runtime the container makes no outbound calls, unless metered licensing is enabled.
• Telemetry: none.

docker run --rm -i \
  -v /srv/documents:/data:ro \
  -v /srv/search-index:/index \
  -v /srv/licenses:/license:ro \
  -e GROUPDOCS_MCP_INDEX_PATH=/index \
  -e GROUPDOCS_LICENSE_PATH=/license/GroupDocs.Search.lic \
  ghcr.io/groupdocs-search/search-net-mcp:26.10.1

• Documents read-only, the index on its own volume — the container can write nowhere else.
• Pin the tag (:26.10.1, not :latest) so a release cannot change behaviour under a running workflow.
• Mount deliberately: the container sees the folders you give it and nothing else — a tighter boundary than a process running as your user.
• On Linux, run with --user matching the owner of the index volume.

Search produces document content as text — file names, matching passages — and that text goes into the conversation. With a cloud-hosted model, “the Acme agreement charges a termination penalty of three monthly fees” has reached your model provider: the document did not move, but the passage did.

For sensitive corpora, that is the argument for a locally-hosted model. Otherwise it is worth knowing, and worth phrasing prompts accordingly: “tell me which files mention the penalty, not the clause” is a request an agent will honour — search returns file names and counts without any document text.

• License file — read from local disk by the local container. Fully offline; the right answer for air-gapped deployments.
• Metered (pay-per-use) — reports usage to GroupDocs servers, so it needs outbound egress. Document content is never part of that report.

Both: Licensing.

A good fit: contract, policy and correspondence search inside a network that does not allow uploads; e-discovery-style questions over local archives; documentation search for engineering teams.

Not what this is: a semantic search service or an OCR engine. It matches words, opens 80+ formats including email archives, and quotes passages; matching by meaning and reading scans are steps you add.

Does any document content leave the machine? Not from the server. Passages the agent quotes travel in the conversation to your model provider.

Does it need internet at runtime? No — only to pull the image, and when metered licensing is enabled.

Can I run it air-gapped? Yes: pre-pull the image, use a license file, pin the tag.

What ports does it open? None. stdio only.

What does the index contain? The documents’ extracted text and search data. Protect and dispose of it like the documents.