# How to watermark documents with AI agents using MCP Add text and image watermarks to documents with an AI agent over MCP: locally, in any of 50+ formats, with the original file left untouched. Watermarking through an agent is the small task nobody schedules and everybody needs: mark this before it goes out, brand these before the client sees them, stamp DRAFT until it is approved. {{< alert style="info" >}} The commands and config snippets on this page are for the **.NET** build of the server — the only platform available today. Installation and client setup: [MCP server for .NET]({{< ref "watermark/net/mcp/_index.md" >}}). Other platforms will expose the same tools with their own launch command; everything else on this page applies unchanged. {{< /alert >}} ## The pattern 1. Put the document in the storage folder the server can see. 2. Ask: *"Add a diagonal CONFIDENTIAL watermark to contract.pdf."* 3. The agent calls [`add_watermark`]({{< ref "watermark/mcp/tools-reference/add-watermark.md" >}}) with the text, a font size, and a rotation. 4. A watermarked copy appears in your output folder; the original is untouched. ## Text or image | You want | Tool | Key parameters | |---|---|---| | CONFIDENTIAL / DRAFT / a date | [`add_watermark`]({{< ref "watermark/mcp/tools-reference/add-watermark.md" >}}) | `text`, `fontSize`, `rotation` | | A logo, stamp, or signature scan | [`add_image_watermark`]({{< ref "watermark/mcp/tools-reference/add-image-watermark.md" >}}) | `watermarkImage`, `opacity`, `rotation` | `rotation: -45` gives the classic diagonal. For image watermarks, `opacity` between 0.2 and 0.4 usually reads as branding rather than as an obstruction. ## Chaining Every call writes a **new file**. To put a logo *and* a text mark on the same document, the second call must take the file the first one produced: > Add the logo at 30% opacity, then add a DRAFT text watermark to that result. Without "to that result", the second call starts from the original and only the DRAFT survives. ## Check the licence before a batch Unlicensed output carries evaluation limitations — which on a watermarking tool means the copies you were about to distribute are not distributable. One call to [`get_license_status`]({{< ref "watermark/mcp/tools-reference/get-license-status.md" >}}) settles it; see [Licensing]({{< ref "watermark/mcp/getting-started/licensing.md" >}}). ## Setup ```bash dnx GroupDocs.Watermark.Mcp --yes ``` with `GROUPDOCS_MCP_STORAGE_PATH` pointing at your documents folder — [per-client config]({{< ref "watermark/net/mcp/install-in-ai-clients.md" >}}) or the [installer]({{< ref "watermark/mcp/getting-started/_index.md" >}}). ## Where to go next * [Batch-watermark a folder]({{< ref "watermark/mcp/use-cases/batch-watermark-a-folder.md" >}}) — the whole distribution set in one prompt. * [Find and remove watermarks]({{< ref "watermark/mcp/use-cases/find-and-remove-watermarks.md" >}}) — and what cannot be removed. * [Brand documents with a logo]({{< ref "watermark/mcp/use-cases/brand-documents-with-a-logo.md" >}}) — opacity, rotation, and taste. * [On-premise architecture]({{< ref "watermark/mcp/use-cases/on-premise-watermarking.md" >}}) — nothing leaves the machine. ## Read more * [3 ways to label AI-generated documents with watermarks via MCP](https://blog.groupdocs.com/watermark/3-ways-watermark-documents-ai-agents-mcp/) — the same scenario on the GroupDocs blog, with worked examples and the trade-offs.