Labsco
MCP SERVER

Atlassian Confluence MCP Server

by anuragagrawal0430 ↗

47 Confluence tools across Cloud, Server and Data Center — pages, spaces, CQL search, labels, comments, attachments, versions, watchers and restrictions.

Team Wikis & Documentation Platforms
Summary
Safe by default, wide once you open it.

Two switches decide what an agent can do on your wiki: writes are off until `CONFLUENCE_READ_ONLY` is false, and deletes and permission changes stay off after that until enabled separately. There is also a tool allowlist and a space allowlist if on/off is too coarse. Worth setting deliberately, because a confused agent can do a lot of quiet damage to a wiki.

What it is

An MCP server for Confluence that talks to Cloud, on-premise Server and Data Center from one package. It goes well past reading pages: spaces get created and deleted, pages get copied and moved, restrictions get set, inline tasks get extracted, export URLs get generated. It ships read-only and you opt into writes.

What you get
  • Page work — read, create, update, patch, append, replace a range, copy, move and delete — `confluence_get_page`, `confluence_create_page`, `confluence_update_page`, `confluence_patch_page`, `confluence_append_to_page`, `confluence_replace_page_range`, `confluence_copy_page`, `confluence_move_page`, `confluence_delete_page`
  • Finding pages: full CQL and text search, lookup by title or by a pasted Confluence URL, and recently modified — `confluence_search`, `confluence_search_pages`, `confluence_get_page_by_title`, `confluence_get_page_by_url`, `confluence_get_recently_modified`
  • Long pages read in pieces rather than all at once — `confluence_get_page_body_chunk`
  • Spaces: list, inspect, find the homepage, create — including a private one — and delete — `confluence_get_spaces`, `confluence_get_space_by_key`, `confluence_get_space_homepage`, `confluence_create_space`, `confluence_create_private_space`, `confluence_delete_space`
  • Hierarchy and history: children, ancestors, the version list and one version's content — `confluence_get_child_pages`, `confluence_get_page_ancestors`, `confluence_get_page_versions`, `confluence_get_page_version`
  • Labels, comments and attachments, including uploading a file from the host — `confluence_add_page_label`, `confluence_delete_page_label`, `confluence_get_page_comments`, `confluence_add_page_comment`, `confluence_get_page_attachments`, `confluence_upload_attachment`
  • Watchers, restrictions and inline tasks, plus PDF and Word export URLs — `confluence_get_page_watchers`, `confluence_watch_page`, `confluence_unwatch_page`, `confluence_get_page_permissions`, `confluence_set_page_permissions`, `confluence_get_page_tasks`, `confluence_export_page`
Requirements

Node.js 18+ and a Confluence instance — Cloud, or Server/Data Center 7.9+. `npx atlassian-confluence-mcp-server` runs it with nothing installed (1.6.0 in package.json). Credentials come from the environment: `CONFLUENCE_BASE_URL` always, then `PAT` alone for Server and Data Center, or `CONFLUENCE_USERNAME` plus `PAT` for Cloud, with `CONFLUENCE_PASSWORD` as the Basic Auth fallback. `CONFLUENCE_READ_ONLY` defaults to `true`, so mutating tools stay off until you set it to `false`, and `CONFLUENCE_ENABLE_DESTRUCTIVE_TOOLS` defaults to `false`, gating deletes and permission changes separately. Attachment uploads stay disabled until `CONFLUENCE_UPLOAD_ALLOWED_DIRS` names the directories they may read from.

Setup effort

One command plus a key — npx atlassian-confluence-mcp-server, then supply credentials