The plugin is for Confluence Data Center. It does not connect to Confluence Cloud or provide Jira tools.
Available tools
The app supports page editing, comments, labels, history, attachments, navigation and page lifecycle operations. Refresh tool discovery after changing the connection; tools/list lists the operations available to the caller under native permissions and administrator policies.
Confluence packages
Use the package for your Confluence major version: confluence9.jar for Confluence 9 or confluence10.jar for Confluence 10. Packages are not interchangeable. The Confluence 9 package uses the Java 17 baseline; the Confluence 10 package requires Java 21. Follow Confluence’s own runtime requirements for your deployment.
Validate the selected client, runtime, proxy/SSO configuration, capacity and cluster behaviour for your deployment. Platform packages do not establish compatibility with every environment.
MCP clients
Clients need the connection contract: remote Streamable HTTP/JSON, caller authentication and a supported protocol version. No everyday AI client currently has an end-to-end compatibility claim documented here. See Unverified MCP client guidance.
Functional limits
-
The app supports page editing, comments, labels, history reads, attachments, navigation, page copying, same-space moves, leaf-page trash and restoration. Permission administration, restoring a historical page revision and permanent purge are not exposed.
-
Related-content operations require access to the current page; selected historical versions can be read. Storage output is bounded at 50,000 characters with explicit truncation; truncated body reads have no continuation.
-
Writes require a complete storage-format body of at most 50,000 characters. Updates need the current page version and replace the complete title/body; there is no automatic conflict merge or idempotent create guarantee.
-
Search returns bounded, permission-filtered batches rather than a guaranteed complete inventory. Query and pagination limits are in Tools | confluence_search.
Deployment limits
JSON-RPC requests are limited to 65,536 characters and authenticated MCP bodies to 196,608 bytes. OAuth token/revocation forms require Content-Length of at most 196,608 bytes (413 above the limit; 411 without a known length). The app applies per-user request-rate limits and streaming limits. Configure suitable proxy timeouts and capacity controls. See Developer and technical integration for proxy setup.
Use Confluence and ingress controls appropriate to the deployment. Production cluster availability, capacity and saved-policy propagation have not been established.
Deployment compatibility
Use a client that supports the app’s Streamable HTTP connection and authentication contract. Verify HTTPS certificate trust, OAuth or PAT configuration, proxies/SSO and native caller permissions in your environment. A successful connection or package installation alone does not establish production capacity or cluster behaviour.
Use node affinity for pending transfers and continuation cursors. Validate failover, policy propagation and workload limits for your own deployment. Client-specific setup guidance is separate from an end-to-end acceptance claim.
Additional tool limits
Lists return limited batches; concurrent edits can change results between requests. Inline attachment operations allow 32 KiB; larger-file transfers use the limits below. Section editing requires top-level headings and a complete current page within 50,000 characters. Footer comments can be changed; inline comment changes are unavailable.
Navigation, larger files and page lifecycle
File transfers use 32 KiB chunks and a 10 MiB default file limit; page/tree copies are bounded to 25 pages and five levels, moves stay within one space, and trash is reversible and limited to pages without children. Restricted source copies require native Confluence. Use the tool catalog and integration reference for exact contracts.
Customer compatibility depends on the selected client and deployment configuration. Confirm proxy/SSO behaviour, multi-node operation and native-editor interactions for your intended use; do not infer them from tool discovery alone.