Doconda

What the Doconda MCP server does

The Doconda MCP server connects an AI agent to Doconda’s document-processing service. It supports four document categories: Word, PowerPoint, Excel, and PDF. Agents can create new files, upload existing files, request reviews, apply edits, and inspect document-related data.

The workflow is designed for document output that can be delivered for further use. Doconda provides editable files, a reviewed version, a PDF, and a preview of each page. This makes the server suitable for agents that need to produce office documents and verify the result before delivery.

How it works

Doconda exposes its capabilities as MCP tools. An MCP-compatible client or agent framework connects to the server, and the model invokes the relevant tool for the requested document task. Available tools include create_document, upload_file, review_document, edit_document, list_documents, get_document, get_report, and read_artifact_data.

The service can be used in two ways. The hosted endpoint at https://api.eu.doconda.com/mcp uses Streamable HTTP and expects an Authorization: Bearer ... header. It can be called by an MCP-capable agent framework or model API, including the OpenAI Responses API. The local option runs through stdio, which is appropriate for clients such as Claude Desktop.

The Doconda MCP server also supports language selection for the remote service through the Accept-Language header. Error messages and report text are in English by default; sending a supported language value such as es requests Spanish responses from the remote server.

Setup and configuration

For local use, configure an MCP client to run npx -y @doconda/mcp and provide the API key through the DOCONDA_API_KEY environment variable. The README shows this arrangement for Claude Desktop. This approach keeps the server process local while it communicates with Doconda using the supplied key.

For remote use, no package installation is required. Configure the client or application with the hosted MCP URL and pass the API key as a bearer token in the request headers. API keys are obtained through Doconda. Applications that do not use MCP can use the official JavaScript or Python SDKs, which expose the same tools for models that support tool calling.

Tools and capabilities

• Create documents with create_document.
• Upload source files with upload_file.
• Review files with review_document and retrieve review information with get_report.
• Apply changes with edit_document.
• Browse and retrieve stored documents with list_documents and get_document.
• Inspect artifact data with read_artifact_data.

Limitations and notes

The supplied material does not specify individual tool parameters, supported file extensions beyond the four document categories, pricing, quotas, or review criteria. Those details should be confirmed in Doconda’s documentation before building a production workflow.

Remote requests require bearer-token authentication, while local configuration requires DOCONDA_API_KEY. The documented local setup uses stdio, and the documented hosted setup uses Streamable HTTP. Responses from the hosted service can be localized through Accept-Language; the default language for errors and reports is English.