Skip to main content
The Model Context Protocol (MCP) is an open protocol, introduced by Anthropic in November 2024, that defines how an AI application (the client) connects to external services (MCP servers). A server exposes tools, resources and prompts through JSON-RPC 2.0 messages, over stdio locally or over HTTP (Streamable HTTP) remotely. One integration then works with any compatible client.

​
Why it matters when generating documents with AI

  • A document MCP server gives Claude, Cursor, Codex and other clients the ability to create files without writing integration code.
  • Tools are described with JSON Schema, the same format used by tool calling.
  • A local server can read files from disk; a remote one receives them by URL or as base64.
  • Authentication depends on the transport: environment variables locally, HTTP headers remotely.

​
With Doconda

Doconda’s MCP server gives your agent the tools to create, review and edit documents. Locally it runs with npx -y @doconda/mcp and the DOCONDA_API_KEY variable; remotely it is at https://api.eu.doconda.com/mcp with an Authorization header. Setup for each client is in MCP server.