Skip to main content
An AI agent is a system in which a language model decides which steps to take to complete a task and carries them out with tools: searching for information, calling an API, reading or writing files. It works in a loop: the model asks for a tool, gets the result and decides the next step until it is done. Agents are built with tool calling or with MCP servers.

​
Why it matters when generating documents with AI

  • Many agent tasks end in a file: a report in Word, a budget in Excel, a presentation.
  • The agent needs tools that return a checkable result (status, link, warnings), not just text.
  • An agent can check its own result and fix it if the tool explains what went wrong.
  • Files the agent reads can carry hidden instructions; they shouldn’t be obeyed.

​
With Doconda

Doconda connects to your agent through the MCP server, the SDK tools or the REST API. The agent gets tools to create, upload, review, edit and list documents and read their report. When creating, if a source tries to give orders, it is ignored and the report says so (source.instructions, see Create).