Triton AI Docs

MCP connections

Give your AI client tools to search Jira issues and read Confluence pages.

Model Context Protocol (MCP) connects an AI client to external tools. Use MCP Atlassian to connect your client to UC San Diego's Jira and Confluence.

The client starts one local process that serves both products. You configure the connection once per client, independently of its model-provider configuration.

Start here

  1. Install the server and create your credentials.
  2. Add the configuration for your client.
  3. Check both connections with a search and a page read.

What the connection provides

The setup starts with four read-only tools:

ToolPurpose
jira_searchFind issues with Jira Query Language (JQL).
jira_get_issueRead an issue by its key.
confluence_searchFind pages with text or Confluence Query Language (CQL).
confluence_get_pageRead a page by its ID or URL.

Jira and Confluence enforce your account's access. Tool results enter the AI client's context, so use an approved client and model for the information you retrieve.

To enable updates, see Allow selected write tools.

How this setup works

These guides use the open-source MCP Atlassian server with local stdio transport. The client launches the process and communicates through standard input and output. No listening port is required.

The examples pin mcp-atlassian==0.23.0, the version checked for this guide. Client configuration examples cover macOS and native Windows. The Windows and WSL notes explain which paths to use for a Linux-based setup.

On this page