# GroupDocs.Parser MCP Server GroupDocs.Parser MCP server lets AI agents like Claude, Cursor, and Copilot extract text, tables, images, metadata, and barcodes from documents — locally on your machine. **GroupDocs.Parser MCP server** lets AI agents like Claude, Cursor, and Copilot **pull data out of documents** — text, tables, images, metadata, barcodes — from PDF, Office, email, e-book, and image formats, **locally on your machine**. Install with one command: ```bash docker run --rm -i --user "$(id -u):$(id -g)" \ -v $(pwd)/documents:/data \ ghcr.io/groupdocs-parser/parser-net-mcp:latest ``` `--user` runs the container as you, so on Linux it can write into a folder you own (the image runs as a non-root user otherwise). Docker Desktop on Windows and macOS does not need it. {{< alert style="info" >}} **This product ships via Docker only.** GroupDocs.Parser's engine embeds around 234 MB of ONNX models, which puts the packed tool over nuget.org's 250 MB package limit — so this product ships **via Docker only**, and there is no `dnx` command for it. Other GroupDocs MCP servers offer a NuGet (`dnx`) channel; this one does not, and every command in these pages reflects that. Details: [MCP server for .NET]({{< ref "parser/net/mcp/_index.md" >}}). {{< /alert >}} Or use the [guided installer]({{< ref "parser/mcp/getting-started/_index.md" >}}) to register the server in your AI client, verify the setup, and configure shared folders in one pass. ## What you can do Seven tools (full details in the [tools reference]({{< ref "parser/mcp/tools-reference/_index.md" >}})): * **[`extract_text`]({{< ref "parser/mcp/tools-reference/extract-text.md" >}})** — plain text, whole document or one page. * **[`extract_tables`]({{< ref "parser/mcp/tools-reference/extract-tables.md" >}})** — tables as Markdown that renders in chat, or JSON rows to compute with. * **[`extract_metadata`]({{< ref "parser/mcp/tools-reference/extract-metadata.md" >}})** — author, title, dates, custom properties, EXIF/XMP/IPTC. * **[`extract_images`]({{< ref "parser/mcp/tools-reference/extract-images.md" >}})** — embedded images saved to storage as files. * **[`extract_barcodes`]({{< ref "parser/mcp/tools-reference/extract-barcodes.md" >}})** — Code128, QR, PDF417, DataMatrix, EAN, UPC, Aztec and more — **including on scans**. * **[`get_document_info`]({{< ref "parser/mcp/tools-reference/get-document-info.md" >}})** — type, pages, size, without extracting. * **[`get_license_status`]({{< ref "parser/mcp/tools-reference/get-license-status.md" >}})** — active licensing mode and metered consumption. Ask in plain language — *"what does this invoice say, and what is in its table?"* — and the agent picks the tools. ## Install for your platform 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** (Docker only) | [MCP server for .NET]({{< ref "parser/net/mcp/_index.md" >}}) | | Java | Planned | [Tell us you need it](https://forum.groupdocs.com/c/parser/17) | | Python | Planned | [Tell us you need it](https://forum.groupdocs.com/c/parser/17) | | Node.js | Planned | [Tell us you need it](https://forum.groupdocs.com/c/parser/17) | ## Extraction is not conversion The distinction that decides which GroupDocs server you want: | You want | Server | |---|---| | Values **out of** a document — text, a table, a barcode, properties | **this one** | | The document **as another format**, layout preserved | [GroupDocs.Conversion](https://docs.groupdocs.com/conversion/mcp.md) | | Clean **Markdown** of the whole document for RAG | [GroupDocs.Markdown](https://docs.groupdocs.com/markdown/mcp.md) | | Metadata **edited or removed**, not just read | [GroupDocs.Metadata](https://docs.groupdocs.com/metadata/mcp.md) | Plenty of pipelines use two: convert for ingestion, parse for extraction. ## Barcodes work where text does not A scanned page has no text layer, so [`extract_text`]({{< ref "parser/mcp/tools-reference/extract-text.md" >}}) returns little — but the engine's models detect **codes in the rasterized image**, so [`extract_barcodes`]({{< ref "parser/mcp/tools-reference/extract-barcodes.md" >}}) still returns the tracking number on a scanned delivery note. That capability is also why the image is large: the models ship inside it. There is **no OCR**: printed words in a scan stay pixels. ## Supported AI clients | 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]({{< ref "parser/net/mcp/install-in-ai-clients.md" >}}). ## Delivery channel | | Docker (the only channel) | |---|---| | Prerequisites | Docker. No .NET SDK — the runtime is in the image | | Native dependencies | bundled in the image | | Image | `ghcr.io/groupdocs-parser/parser-net-mcp` · `groupdocs/parser-net-mcp` | | Architectures | linux/amd64 + linux/arm64 (Apple Silicon native) | | Note | the image is large (it carries the ONNX models) — pull it once, deliberately | ## How it works The server uses MCP's **local stdio transport**: your AI client starts the container 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*. Extraction is where document content most obviously leaves a building; here it does not. Details: [On-premise architecture]({{< ref "parser/mcp/use-cases/on-premise-document-extraction.md" >}}). ## When you need more than a text-extraction script Scripts that pull text from PDFs are easy to find and hard to trust across formats. Choose this server when you need: **one interface across 50+ formats** including email and e-books; **tables as tables**, not as a soup of positioned strings; barcode decoding that works on scans; metadata across documents and images in the same call; and the fidelity of the commercial GroupDocs engine trusted by enterprise teams for over a decade. ## Resources * [Quick start]({{< ref "parser/mcp/getting-started/_index.md" >}}) · [Use cases]({{< ref "parser/mcp/use-cases/_index.md" >}}) · [Troubleshooting & FAQ]({{< ref "parser/mcp/troubleshooting-faq.md" >}}) * GitHub: [server source](https://github.com/groupdocs-parser/GroupDocs.Parser.Mcp) · [installer](https://github.com/groupdocs/GroupDocs.Mcp.Installer) · [integration tests](https://github.com/groupdocs-parser/GroupDocs.Parser.Mcp.Tests) * [Docker image](https://github.com/orgs/groupdocs-parser/packages/container/package/parser-net-mcp) · [MCP Registry](https://registry.modelcontextprotocol.io/v0/servers?search=io.github.groupdocs-parser/groupdocs-parser-mcp) * Questions: [Parser forum](https://forum.groupdocs.com/c/parser/17)