Jira, Confluence, and Compass with the Atlassian Rovo MCP server in meinGPT
The Atlassian integration connects your chat or assistant directly to Jira, Confluence, and Compass so you can search, read, and edit tickets, project context, and knowledge pages — without leaving meinGPT.
Search and read Confluence pages and spaces, and create or update pages
Summarize content from Jira and Confluence
Query Compass components
All tool calls run in the context of the authenticated user — your Jira, Confluence, and Compass permissions remain in effect.
Note
Jira and Confluence used to be separate connectors. They are now combined into the
Atlassian connector. Existing assistants are migrated automatically — you only
need to reconnect once.
The Atlassian integration uses OAuth 2.1 with dynamic client registration — you don't need to create an OAuth app or store an API token manually.
Enable the connector: Select Atlassian among the connectors in the assistant editor and save the assistant — or enable Atlassian directly in a chat via the "+" menu.
Connect: Under "Connections" (or on the first tool call in the chat), click "Connect". An Atlassian window opens for authorization.
Confirm access: Choose your Atlassian site and confirm the requested permissions.
Done: After authorization, Atlassian shows as connected under "Connections" and is available in the chat.
If your Atlassian organization restricts which domains may connect via the Rovo MCP server settings (for example with Atlassian Guard), connecting shows a message like "Your organization admin must authorize access from this redirect URL to this site". An Atlassian organization admin then authorizes the meinGPT domain once:
Open Atlassian Administration: Go to admin.atlassian.com and select your organization.
Rovo MCP server settings: Open Rovo → Rovo MCP server → Domains.
Add domain: Click Add domain and enter your meinGPT instance origin including protocol and callback path, for example https://app.meingpt.com/oauth/callback (for on-premise, your own instance URL).
Save and click "Connect" again in meinGPT.
Note
Authorizing the domain is organization-wide and one-time, and connects no
users — it only removes the blocker. The connection itself is per user:
each person clicks "Connect" once so tool calls run with their own Atlassian
permissions.
"Show me all open issues in the AUTILITY project assigned to me"
"Create an issue in AUTILITY: backlog import fails on empty fields"
"Summarize the current state of ticket AUTILITY-142"
"Search the Engineering space for pages about onboarding"
"Read the 'Architecture overview' page and summarize the key points"
"Create a page in the Product space with today's meeting notes"
The connector is built for permission-sensitive operational queries in user context. For enterprise-wide search across large knowledge estates, you can additionally ingest Confluence content into a data pool (for example via a Outpost source) and search it at scale.
Connector: permission-sensitive operational queries per user
Data pool: scalable search and RAG across large estates
Important: the connector's user-level permissions do not carry over to the data pool automatically