Troubleshooting & FAQ Leave feedback

Solutions to the most common GroupDocs.Redaction MCP server issues — server not appearing in the client, startup failures, missing native dependencies, and first-launch timeouts.

Note

Platform-specific troubleshooting: runtime problems depend on which build you run. For the dnx runner, native graphics libraries, and the Docker channel, see Troubleshooting (.NET). The issues on this page apply to every platform.

1. Restart the client — every client reads its MCP config only at startup.
2. Check the config file location for your OS (per-client reference) and that the entry sits under the right root key (mcpServers for Claude Desktop/Cursor/Windsurf, servers for VS Code/VS 2022).
3. Validate the JSON — a trailing comma silently breaks the whole file. If you used the installer, a timestamped .bak of your previous config sits next to the file for comparison.

A cold cache: on the very first use the server’s package or image is still downloading while the client is already waiting on the connection. Warming it once fixes it for good — the exact command depends on your build: .NET.

These are properties of the build you run rather than of MCP, so the fixes live with the platform:

Pass the file name, not a full path from your machine: the server resolves names inside its configured storage folder. When a name is not found the tool responds with the list of files it can see, so the agent can correct itself — check that list against GROUPDOCS_MCP_STORAGE_PATH.

For redact_text, yes — the matching text is replaced in the document rather than covered over, which is the difference between redaction and drawing a black rectangle in a viewer. But “really gone” is a claim about the whole file, so finish the job: text can also live in annotations, in metadata, and in images of the page. That is why this server has four redaction tools rather than one.

No. You (or the agent) supply the pattern. The tools apply it exactly and predictably — there is no classifier guessing at what looks like a national ID. That is a deliberate boundary: an agent proposing patterns is useful; an agent silently deciding what counts as sensitive is not something to trust with a legal disclosure.

redact_text works on the text layer. A name that appears in a scanned page is part of an image, and must be covered with redact_image_area using coordinates. There is no OCR step in this server.

Read the redacted copy back and search it: run the same pattern again and expect zero matches, check the metadata with erase_metadata applied, and look at the pages. The verification workflow sets this out step by step. Do not skip it — a redaction you have not verified is a claim, not a control.

Almost certainly evaluation mode: one redaction per document, capped at four replacements. A pattern matching twelve times leaves eight in place, and nothing in the response says so. get_license_status first, every time.

Ask your agent “list your GroupDocs redaction tools and the license status” — it should name redact_text, redact_image_area, redact_annotations, erase_metadata, get_document_info, get_license_status. For a scripted check that performs the real MCP handshake and a live call through the engine, see verifying a .NET installation.

Post your config (redact license paths) and the client name in the Redaction forum — we answer MCP questions daily. Bugs: GitHub issues.