MCP Server: Generate Documents with AI

What if your AI agent could generate an invoice, produce a compliant PDF or send a document out for signature without ever leaving the conversation? That is exactly what the LayerOne MCP server makes possible. Thanks to the Model Context Protocol, you connect your document and signature tools directly to Claude, ChatGPT or Cursor, and you drive everything with simple instructions in plain language.

What is MCP (Model Context Protocol)?

The Model Context Protocol, or MCP, became the open standard in 2026 for connecting external tools to artificial intelligence agents. Introduced by Anthropic, it has been adopted by the major players in the industry — Anthropic, OpenAI, Google and Microsoft — which makes it today the universal reference for extending the capabilities of an AI assistant.

In practical terms, MCP acts as a "universal port" between the agent and the outside world. Where you previously had to build a custom integration for every single tool, the protocol enforces a common language: an MCP server exposes a set of tools, and any compatible client (Claude Desktop, Cursor, ChatGPT and more) can use them without specific configuration. The AI agent discovers the available tools, understands what they do, and triggers them at the right moment based on what you ask.

What the LayerOne MCP server enables

The LayerOne MCP server, named layerone-mcp, gives your AI agent 20 tools covering two complementary areas: document generation and electronic signature. In short, your assistant can produce a professional document and have it signed, from start to finish, without you ever touching a technical interface.

11 tools to generate documents and invoices

The DocX side exposes 11 document generation tools. Your agent can create PDFs from templates, fill in fields, and produce electronic invoices in the Factur-X format, the Franco-German standard compliant with French regulations. The document engine is hosted at docx.layerone.fr.

9 tools for electronic signature

The Sign side brings 9 electronic signature tools compliant with the European eIDAS regulation. Your agent can send a document out for signature, track signing progress and retrieve the final signed document. The signature service is hosted at sign.layerone.fr.

How to install the LayerOne MCP server

The layerone-mcp server is published on npm and registered in the official MCP registry under the identifier io.github.GOLayerone/layerone. Installation requires no build step: the npx -y layerone-mcp command downloads and launches the server automatically.

All you need to do is add the following configuration to your MCP client's settings file (Claude Desktop, Cursor or ChatGPT):

{
  "mcpServers": {
    "layerone": {
      "command": "npx",
      "args": ["-y", "layerone-mcp"],
      "env": {
        "DOCX_API_KEY": "your_docx_key",
        "SIGN_API_KEY": "your_sign_key"
      }
    }
  }
}

The two environment variables DOCX_API_KEY and SIGN_API_KEY hold your access keys. You get them in a few clicks on the developer portal dev.layerone.fr. The free tier lets you generate up to 20 documents per month, enough to try everything with no commitment. Once the configuration is saved and your client restarted, the 20 LayerOne tools appear automatically in your conversation.

Real-world use cases

Once the server is connected, everything is driven in plain language. A few examples of what you can ask your assistant:

• "Generate the invoice for client X and send it for signature": the agent creates the Factur-X invoice, then triggers the electronic signature request, in a single instruction.
• "Produce the quote for this project as a PDF": the agent fills in the template and returns the document ready to send.
• "What is the status of the contract I sent yesterday?": the agent queries the Sign service and gives you the status in real time.
• "Retrieve the signed PDF of the latest quote": the agent downloads the final document along with its signature proof.

For a developer working in Cursor, this means generating test documents or invoices directly from the editor. For a Claude Desktop or ChatGPT user, it means automating administrative paperwork without ever opening an invoicing app.

Why use MCP rather than a classic API?

You can, of course, use the LayerOne REST API directly in your code. But the MCP server adds an extra layer: the AI agent becomes able to decide on its own which tool to call, and in what order, based on your request. You no longer write integration code — you describe the result you want, and the assistant orchestrates the calls. It is the ideal approach for automating repetitive business tasks without heavy development.

Which AI clients are compatible?

The LayerOne MCP server follows the official Model Context Protocol standard, so it works with any MCP-compatible client. The most widely used in 2026 are Claude Desktop, Anthropic's desktop application, Cursor, the AI-driven code editor popular with developers, and ChatGPT, which also supports MCP servers. The mcpServers configuration shown above is identical for all three: only the location of the settings file changes depending on the client. By design, any new tool that adopts the protocol will be able to use the 20 LayerOne tools without any adaptation on your part.

Security and privacy

Your DOCX_API_KEY and SIGN_API_KEY keys stay stored locally in your MCP client's configuration: they never pass through a third party. Every tool call is authenticated against the LayerOne services, and the documents generated as well as the signatures carry the same guarantees as the LayerOne API — Factur-X for French tax compliance and eIDAS for the legal value of signatures. You keep full control over your data while delegating the mechanics to your AI agent.

Get started in minutes

To use the LayerOne MCP server, create a free account, grab your API keys and add the configuration to your MCP client. In under five minutes, your AI agent knows how to generate compliant documents and launch electronic signatures.

Frequently asked questions

What is MCP?

MCP stands for Model Context Protocol, the open standard for connecting external tools to AI agents. Introduced by Anthropic and adopted by Anthropic, OpenAI, Google and Microsoft, it lets an MCP server expose a set of tools that any compatible client — Claude Desktop, Cursor or ChatGPT — can use without custom integration.

Is the LayerOne MCP server free?

Yes, to start. The free tier lets you generate up to 20 documents per month with no commitment, enough to try every tool the layerone-mcp server exposes.

How do I get the API keys for the MCP server?

You get your DOCX_API_KEY and SIGN_API_KEY in a few clicks on the developer portal at dev.layerone.fr, then add them to the env section of your MCP client's configuration.

Which AI clients are compatible?

Any MCP-compatible client. The most widely used in 2026 are Claude Desktop, Cursor and ChatGPT. The mcpServers configuration is identical for all three — only the location of the settings file changes.

Related articles