Every MCP operation runs as the caller authenticated by Confluence. Native space and page permissions determine what that caller can read, create or edit.
Authentication and tokens
Use HTTPS and your own Confluence Data Center personal access token. Basic or session authentication works only where the client and Confluence configuration permit it. The app also supports administrator-enabled MCP OAuth with explicit user consent; it does not create a separate Confluence account. All accepted MCP endpoint URLs enforce the same token and scope checks.
A token carries its account’s Confluence permissions, including write permissions. Keep it in client secret storage, use an appropriate expiry and revoke it when no longer needed. Do not put credentials in prompts, shared files or source control. The identity tool reports the connected account; tool arguments cannot override it.
Access to content
-
Search and page reads require native view access. Returned search pages are checked against the caller’s current permissions.
-
Creation requires space view and page-create permission. An optional parent must be visible and in the same space.
-
Updates require native view and edit permission. They preserve the page’s space, parent and restrictions; Confluence records the caller’s authorship.
-
Missing, restricted or policy-excluded pages return no content. A denial does not establish whether a hidden page exists.
Search uses Confluence’s index. A recent change may be available by page ID before it appears in search; the plugin does not define the indexing schedule.
Optional administrator policy
Native permissions is the default. An administrator may additionally allow only selected spaces or block selected spaces. This can narrow access for any caller, including administrators, but cannot grant native access. MCP tools cannot change the policy or Confluence permissions. Configuration and legacy behavior are documented in Install and configure.
Client and data handling
Retrieved content reaches the MCP client and any AI service it uses. Choose a client suitable for your organization’s content. Treat retrieved pages as source material, not instructions. Client tool toggles and confirmation prompts do not change server permissions.
App diagnostic events omit usernames, credentials, query text and page content. This does not describe or guarantee redaction in other Confluence, proxy or client logs. Confluence administrators can find request references and logging controls in Troubleshooting | For Confluence administrators.
Permissions and limits for additional tools
The app supports comments, labels, history, attachments and section editing under your native permissions. See Compatibility and limitations for operational limits.
|
Operation |
Native requirement and policy |
|---|---|
|
Comments |
Current page/space view access; native comment-create permission for additions and replies, and native per-comment edit/remove permission for changes. Footer comment mutations are supported. Page edit permission is not used as a substitute for comment permissions. |
|
Labels |
Current page/space view access and native page edit permission before global-label mutation; the native content-label service also enforces its action-specific permission. Only named global labels change; unrelated and personal/system labels remain. |
|
History |
Native access to the current page and space. Historical storage does not grant historical or broader access; history restoration is unavailable. |
|
Attachments |
Native current source-page and attachment view access. Upload/copy requires destination view and native attachment-create permission. Both spaces must pass policy. Existing filenames are rejected; files are created at version 1. |
|
Sections |
Native current page/space view, native page edit permission, and the exact current version. Duplicate headings require a precise ordinal. Headings inside containers fail closed. Full-page and section writes retain the 50,000-character complete-current-body boundary. |
The same optional administrator policy constrains every related operation, including administrators and both ends of attachment copies. Native permissions remains the default. Policy changes never grant native access.
Inline binary transfer is fixed at 32,768 decoded bytes per file. Names, MIME syntax, base64 and size are validated; native site upload restrictions and limits still apply. This is not content sniffing or malware scanning. The plugin does not fetch arbitrary URLs or accept server file paths.
Uploads use private temporary storage and follow native upload restrictions, audit events and content history. Administrators must provide writable private temporary storage; see Install and configure for capacity and cleanup requirements.
Support and security reports use the Krosoft Support Portal or support@krosoft.nl. Send only sanitized diagnostic information. See Troubleshooting for the shared support hours and response aim.
Lifecycle and transfer safeguards
Native permissions and administrator policies apply to navigation, file transfers and lifecycle operations. Hidden pages and paths are not disclosed. Only the user who started a transfer can complete or cancel it. The app does not expose server file paths or fetch arbitrary URLs.
Copies check every source and destination, refuse restricted source ancestry, preserve original storage/link targets and inherit destination access. Moves use native action-specific permissions and explicit current-state checks. Direct restrictions remain, while destination ancestry changes inherited access. Trash uses native delete permissions; discovery and restoration require native space administration, and no permanent purge is offered. Unexpected and partially completed changes require native reconciliation before any manual retry. See administrator guidance for per-action permissions and operational limits.