# How to assemble a document pack with an AI agent Assemble a multi-part document pack with an AI agent over MCP: merge in rounds, keep the order, and produce one file for distribution. Tender packs, board papers, case bundles, onboarding packets: the same shape every time — a fixed order, more than four parts, and a deadline. This is the workflow. {{< 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 "merger/net/mcp/_index.md" >}}). Other platforms will expose the same tools with their own launch command; everything else on this page applies unchanged. {{< /alert >}} ## Name the files so order is obvious ```text D:/Storage/Documents/ 00-cover.pdf 01-summary.pdf 02-body.pdf 03-appendix-a.pdf 04-appendix-b.pdf 05-annexes.pdf ``` A numeric prefix removes all ambiguity about sequence — for you and for the agent. Before a repeat run, set `GROUPDOCS_MCP_OUTPUT_PATH` to a separate folder ([configuration]({{< ref "merger/net/mcp/configuration.md" >}})). Outputs otherwise land in the storage folder, where the next *"every PDF in my documents folder"* run would merge last month's pack and the intermediate results back in. ## The prompt > Merge every PDF in my documents folder into one pack, in file-name order. Merge four at a time, chaining each result into the next call, and tell me the final file name. The agent runs [`merge`]({{< ref "merger/mcp/tools-reference/merge.md" >}}) in rounds: 00-03 → result, result + 04-05 → final. Two calls here; more parts simply mean more rounds. ## Verify before you send > How many pages does the final pack have? [`get_document_info`]({{< ref "merger/mcp/tools-reference/get-document-info.md" >}}) on the result, compared against the sum of the parts, is a two-second check that catches the two failure modes that matter: a part silently missing, and the **three-page evaluation trim**. A pack that should be 87 pages and reports 3 was produced unlicensed — [`get_license_status`]({{< ref "merger/mcp/tools-reference/get-license-status.md" >}}). ## When the parts are not all PDFs Convert first, then assemble: > Convert the Word sections to PDF, then merge everything in file-name order. With both the [Conversion](https://docs.groupdocs.com/conversion/mcp.md) and Merger servers registered, that is one conversation. ## Finishing touches other servers provide A pack usually needs more than assembly: * **Mark it** — a CONFIDENTIAL or DRAFT watermark: [GroupDocs.Watermark](https://docs.groupdocs.com/watermark/mcp.md). * **Clean it** — strip metadata before it leaves: [GroupDocs.Metadata](https://docs.groupdocs.com/metadata/mcp.md). * **Sign it** — a digital signature on the final file: [GroupDocs.Signature](https://docs.groupdocs.com/signature/mcp.md). Apply it last: any later change, including a watermark or a metadata strip, invalidates it. Each is a separate server with the same install pattern, and an agent that has them all can run the whole sequence from one prompt — locally, in the right order, with the file names it produced at each step. ## Read more * [Enforce document pack assembly order with automated AI workflows using MCP](https://blog.groupdocs.com/merger/enforce-document-pack-assembly-automated-mcp/) — the same scenario on the GroupDocs blog, with worked examples and the trade-offs.