How to Build an AI Agent to Chat with Confluence Documentation

[#1]

Hi,

In this video How to build Confluence Documentation Agent, I show you how to build an AI assistant for Confluence documentation using Google ADK and Atlassian MCP — a simple but powerful alternative to the classical retrieval augmented generation (RAG) approach. It lets you chat directly with your Confluence-based documentation if embedded within your product.

This agent enables natural language chat with Confluence-based product documentation (or any other information stored in Confluence). Instead of relying on embedding-based search, it reads a Markdown index file (a list of page titles, IDs, and links) once, stores it in the tool context state, and then looks up the most relevant Confluence pages on demand. In this example, it generates accurate answers using the Marmind Knowledge Base in Confluence, while guiding the user to the mirrored page in the publicly available documentation - https://knowledgebase.marmind.com/kb/

Happy to answer any questions or help you build an Agent for your Confluence needs. Enjoy!

[KennFromEesel]

I’m from eesel AI. One approach I’ve found useful is to have the agent index key Confluence pages in a structured way, so it can reference the most relevant sections on demand. This avoids relying solely on embedding searches and keeps responses accurate. Happy to discuss ideas or approaches for handling the context efficiently.