Creativity · MCP — server

MCP Confluence Server

The MCP Confluence server turns your wiki into a first-class tool for LLM clients. Atlassian provides a hosted MCP endpoint for Confluence Cloud alongside the Jira one, and mcp-atlassian is the go-to community implementation for Data Center. You can search CQL, read and summarize pages, and create new pages — excellent for 'draft a runbook for this on-call topic' or 'which pages mention our old billing system?'.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: Python
Transports: stdio, http

Capabilities

Tools: search (CQL), get_page, create_page, update_page, get_comments
Resources: space and page URIs
Auth: OAuth (hosted) or API token + email (self-hosted)

Install

uvx mcp-atlassian

Configuration

{
  "mcpServers": {
    "confluence": {
      "command": "uvx",
      "args": ["mcp-atlassian"],
      "env": {
        "CONFLUENCE_URL": "https://your-org.atlassian.net/wiki",
        "CONFLUENCE_USERNAME": "[email protected]",
        "CONFLUENCE_API_TOKEN": "..."
      }
    }
  }
}

Frequently asked questions

Can the Confluence MCP server do semantic search?

The server exposes Confluence's CQL search, which is lexical/metadata-based. Semantic search happens on the LLM side: it reads the retrieved pages and reasons over them.

Is it safe to let the model edit pages?

Only if the API token is restricted to a bot account with limited space permissions. Otherwise the LLM could overwrite pages it shouldn't. Require tool approvals on edit actions.

Does it handle Confluence Server (not Cloud)?

Yes — mcp-atlassian supports both URLs. Note that some APIs differ slightly between Cloud and Data Center; check the tool catalogue for your version.

Sources

Atlassian Remote MCP — accessed 2026-04-20
mcp-atlassian GitHub — accessed 2026-04-20