Atlassian Remote MCP Server beta now available for desktop applications

[Ueslei Lima]

Hi @Kun Chen  I'm trying to implement an automated workflow using this mcp server, is it possible to use Api Tokens instead of relying on a browser based authentication request?

[Monique vdB]

@Ueslei Lima and anyone else with questions or issues - the best place to get support from the team is to go directly via the support portal - here is the link!

[Espen Rydningen]

Everyone who's interested in getting the MPC connection also for Bitbucket Cloud should vote this feature up: https://jira.atlassian.com/browse/BCLOUD-23748

[Arun Gangadharaiah]

@Kun Chen @Alex Alex @Alex Mansour  Do we have a tool to upload file or images? - https://community.atlassian.com/forums/Atlassian-Platform-questions/can-we-upload-files-using-remote-mcp-server/qaq-p/3078002

[Darryl Lee]

Oh hey, in case you wanted to use MCP to do any complex editing of Confluence pages (like to fix links whose titles are broken by some weird legacy conversion issue), be advised (by Github Copilot + MCP):

Me:

Oh no, the MCP cannot get storage format? It can only update pages using Markdown? That's kind of lame.

Github Copilot GCP-4.1 + MCP:

You're absolutely right - that is quite limiting! The MCP Atlassian integration only works with Markdown format, not the raw Confluence storage format. This means we can't directly access and surgically edit the <ac:link-body> elements in the storage format.

Legacy Confluence content often has complex macros, panels, structured layouts, and other elements that simply don't have Markdown equivalents. When you try to update such pages through the Markdown interface, you lose all that rich formatting and structure.

This is indeed a significant limitation of the MCP Atlassian integration - it's designed for simple content but falls short for complex, legacy Confluence pages that rely heavily on:

• Panel macros (info, warning, note panels)
• Complex layouts and columns
• Custom macros
• Structured formatting
• Legacy content types

Your original Python script approach was actually the right one for this use case. For pages with legacy content, you'd need to:

• Use the direct Confluence REST API to get the raw storage format
• Apply your regex transformation to remove just the <ac:link-body> elements
• Update the page using the REST API with the storage format

The MCP integration is unfortunately too simplified for this kind of surgical editing on complex pages. It's great for basic content management but not for preserving the intricate structure of legacy Confluence pages.

[chenh]

For some reason, my mcp-remote shutdown after making connection running in Python. No issue with running npc -y mcp-remote. Any idea?

2025-08-10T01:48:31.775Z][10364] Token request stack trace: Error

[2025-08-10T01:48:31.775Z][10364] Token result: {
found: true,
hasAccessToken: true,
hasRefreshToken: true,
expiresIn: '3300 seconds',
isExpired: false,
expiresInValue: 3300
}
[10364] [Remote→Local] 2
[2025-08-10T01:48:31.849Z][10364] [Remote→Local] 2
[2025-08-10T01:48:31.850Z][10364] Remote → Local message {
method: undefined,
id: 2,
result: 'result-present',
error: undefined
}

Available MCP Tools: ['atlassianUserInfo', 'getAccessibleAtlassianResources', 'getConfluenceSpaces', 'getConfluencePage', 'getPagesInConfluenceSpace', 'getConfluencePageAncestors', 'getConfluencePageFooterComments', 'getConfluencePageInlineComments', 'getConfluencePageDescendants', 'createConfluencePage', 'updateConfluencePage', 'createConfluenceFooterComment', 'createConfluenceInlineComment', 'searchConfluenceUsingCql', 'getJiraIssue', 'editJiraIssue', 'createJiraIssue', 'getTransitionsForJiraIssue', 'transitionJiraIssue', 'lookupJiraAccountId', 'searchJiraIssuesUsingJql', 'addCommentToJiraIssue', 'getJiraIssueRemoteIssueLinks', 'getVisibleJiraProjects', 'getJiraProjectIssueTypesMetadata', "Read a file's content"]
[10364]
Shutting down...
[2025-08-10T01:48:31.899Z][10364]
Shutting down...
[2025-08-10T01:48:31.901Z][10364] Remote transport closed, closing local transport