# Quick start Run the GroupDocs.Search MCP server with Docker, register it in Claude Desktop, Claude Code, VS Code or Cursor, and ask your first question about a folder of documents — in five minutes. {{< alert style="info" >}} These steps run the **.NET** build of the server, which ships as a Docker image — the only platform available today. Platform-independent material (tools, use cases, licensing) lives here and applies to every future platform. {{< /alert >}} Run the GroupDocs.Search MCP server with Docker, register it in your AI client, and ask your first question about a folder of documents — locally, in five minutes. ## 1. Pull the image ```bash docker pull ghcr.io/groupdocs-search/search-net-mcp:latest ``` About 380 MiB. Pulling it once now means your client does not time out waiting for it on the first call. ## 2. Register it in your AI client **Claude Code:** ```bash claude mcp add groupdocs-search -- docker run --rm -i \ -v /path/to/documents:/data ghcr.io/groupdocs-search/search-net-mcp:latest ``` **Claude Desktop, Cursor, Windsurf** — add under `mcpServers` in the client's config file: ```json { "mcpServers": { "groupdocs-search": { "command": "docker", "args": [ "run", "--rm", "-i", "-v", "/path/to/documents:/data", "ghcr.io/groupdocs-search/search-net-mcp:latest" ] } } } ``` Replace `/path/to/documents` with the folder the agent may search. **On Linux**, add `"--user", "1000:1000"` (your numbers from `id -u` and `id -g`) after `"-i"` — [why]({{< ref "search/net/mcp/linux-installation.md" >}}#run-the-container-as-your-user). VS Code, Visual Studio, Cline, Codex and one-click install links: [Register in AI clients]({{< ref "search/net/mcp/install-in-ai-clients.md" >}}). Then **restart your AI client** so it picks up the new server. ## 3. Your first question Put a folder of documents (say `contracts/`) into the documents folder, then ask: > Index the contracts folder, then tell me which contracts mention a termination penalty and quote the clause The agent calls `add_to_index`, then `search`, then `get_text_fragments`, and answers with the file names and the quoted passages. The index stays on disk, so the next question does not re-read anything. The index is written to `.groupdocs-search-index` inside your documents folder; the server never indexes that folder itself. In a code repository, add it to `.gitignore`. ## Where to go next * **The pattern in detail:** [Search documents with AI agents]({{< ref "search/mcp/use-cases/search-documents-with-ai-agents.md" >}}) * **Your OS:** [Windows]({{< ref "search/net/mcp/windows-installation.md" >}}) · [Linux]({{< ref "search/net/mcp/linux-installation.md" >}}) · [macOS]({{< ref "search/net/mcp/macos-installation.md" >}}) * **Read-only documents, a separate index volume, metered keys:** [Configuration]({{< ref "search/net/mcp/configuration.md" >}}) * **Licensing:** evaluation mode caps an index at 100 documents and a result at 10 — [Licensing]({{< ref "search/mcp/getting-started/licensing.md" >}}).