GroupDocs.Redaction MCP Server Leave feedback
On this page
GroupDocs.Redaction MCP server lets AI agents like Claude, Cursor, and Copilot redact documents — text by pattern, areas of a page, annotations, and metadata — locally on your machine. For a redaction tool that property is not a feature, it is the requirement: the documents you redact are the ones that must not be uploaded anywhere.
Run it with one command. The Docker image is self-contained — the runtime and every native dependency the engine needs are inside it:
docker run --rm -i -v $(pwd)/documents:/data \
ghcr.io/groupdocs-redaction/redaction-net-mcp:latest
With the .NET 10 SDK installed, the same server also runs without Docker:
dnx GroupDocs.Redaction.Mcp --yes
Both are the .NET build of the server and run on Windows, Linux, and macOS. Other platforms will each get their own launcher — see Install for your platform.
Or use the guided installer to register the server in your AI client, verify the setup, and configure shared folders in one pass.
WarningEvaluation mode produces incomplete redactions. One document per process, one redaction, capped at four replacements, plus trial badges — and nothing in the response says a match was skipped. A document redacted unlicensed can look clean and still contain the data. Checkget_license_statusbefore every run that matters; see Licensing.
Sensitive data hides in four places, and there is a tool for each (full details in the tools reference):
| Where the data is | Tool |
|---|---|
| In the text | redact_text — regular-expression matching, replaced in the document |
| In an image or scan | redact_image_area — a solid box over a rectangle |
| In comments and notes | redact_annotations — replace or delete annotations |
| In the file’s properties | erase_metadata — author, company, dates, keywords |
Plus get_document_info for page dimensions and get_license_status for the check above.
A complete pass uses all four. Cleaning the body text and shipping a file whose margin comments and author field still name the person is the most common way redactions fail.
Installation, prerequisites, and client configuration are platform-specific; the tools and licensing model below are the same everywhere.
| Platform | Status | Install and setup |
|---|---|---|
| .NET | Available | MCP server for .NET |
| Java | Planned | Tell us you need it |
| Python | Planned | Tell us you need it |
| Node.js | Planned | Tell us you need it |
The agent proposes patterns and drives the calls; the engine applies them exactly. There is no classifier deciding what looks sensitive — you get predictable, reviewable behaviour, and the responsibility for what counts as sensitive stays with you.
That division is deliberate. An agent that quietly decided which names to remove would be impossible to audit, and a redaction you cannot audit is not a control. Ask the agent to report match counts and to verify afterwards — how to verify.
| Client | How it connects |
|---|---|
| Claude Desktop | claude_desktop_config.json |
| Claude Code | claude mcp add CLI |
| VS Code / GitHub Copilot | user-level or workspace mcp.json |
| Visual Studio 2022 (17.14+) | .mcp.json in the solution root |
| Cursor | ~/.cursor/mcp.json |
| Windsurf | ~/.codeium/windsurf/mcp_config.json |
| Cline | Cline MCP settings |
| Codex CLI | codex mcp add CLI |
| JetBrains Rider | manual registration (Settings → AI Assistant → MCP) |
Exact config blocks for every client: Register in AI clients.
| Docker (recommended) | NuGet (dnx) | |
|---|---|---|
| Prerequisites | Docker only | .NET 10 SDK (+ libgdiplus on Linux/macOS) |
| Native dependencies | bundled in the image | installed by you (or the setup script) |
| Package | ghcr.io/groupdocs-redaction/redaction-net-mcp | GroupDocs.Redaction.Mcp on NuGet |
| Architectures | linux/amd64 + linux/arm64 (Apple Silicon native) | any OS with .NET 10 |
The server uses MCP’s local stdio transport: your AI client starts the server as a child process and talks to it over standard input/output. No inbound ports, no external endpoints, no telemetry — the data path is agent → local server → local filesystem. Details, including what still reaches your model provider: On-premise architecture.
Drawing a box in a PDF viewer leaves the text underneath, and everyone has seen a “redacted” document that could be copy-pasted. Choose this server when you need: text replaced in the document, not covered; the same model across 30+ formats; annotations and metadata handled as part of the same pass; pattern-based work an agent can drive over many files; and the fidelity of the commercial GroupDocs engine trusted by enterprise teams for over a decade.
- Quick start · Use cases · Troubleshooting & FAQ
- GitHub: server source · installer · integration tests
- NuGet package · Docker image · MCP Registry
- Questions: Redaction forum
Was this page helpful?
Any additional feedback you'd like to share with us?
Please tell us how we can improve this page.
Thank you for your feedback!
We value your opinion. Your feedback will help us improve our documentation.
