Confluence Dc MCP

Confluence Data Center — MCP Skill (mcp-atlassian)

This skill governs how the agent interacts with a self-hosted Confluence Data Center (or Server) instance via the mcp-atlassian MCP server. It complements the jira-create-issues skill; both products share one MCP server process.

Architecture

┌──────────────┐  stdio   ┌──────────────────┐  HTTPS/PAT  ┌─────────────────────┐
│ Cursor /     │◄────────►│ mcp-atlassian    │◄───────────►│ Confluence DC       │
│ Claude Code  │          │ (local process)  │             │ (on-prem / VPN)     │
└──────────────┘          └──────────────────┘             └─────────────────────┘

The MCP server runs locally on the user's machine (inside the VPN) and connects to Confluence DC via its REST API using a Personal Access Token (PAT).

MANDATORY: Approval before write operations

NEVER create, update, or delete Confluence pages without the user's explicit approval.

When the user asks to create or update a Confluence page:

1. Draft the content in the conversation (or as a local file) and present it to the user for review.
2. Wait for explicit approval — the user must confirm before calling confluence_create_page, confluence_update_page, or confluence_delete_page.
3. Only then execute the write operation.

This applies to all write operations: creating pages, updating page content, adding comments, uploading attachments, moving pages, adding/removing labels, and deleting pages.

Read operations (confluence_search, confluence_get_page, confluence_get_page_children, etc.) do not require approval.

Environment variables

Tip: If the user already has mcp-atlassian configured for Jira, add the CONFLUENCE_* env vars to the same server entry — both products share one process.

MCP config snippets

Cursor (Settings → AI → MCP Servers)

Claude Code / Claude Desktop (~/.claude/mcp_servers.json)

{
  "mcp-atlassian": {
    "command": "uvx",
    "args": ["mcp-atlassian"],
    "env": {
      "JIRA_URL": "https://jira.yourcompany.com",
      "JIRA_PERSONAL_TOKEN": "your-jira-pat",
      "CONFLUENCE_URL": "https://confluence.yourcompany.com",
      "CONFLUENCE_PERSONAL_TOKEN": "your-confluence-pat"
    }
  }
}

Config file locations:

• Windows (installer): %APPDATA%\Claude\claude_desktop_config.json
• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Generating a Personal Access Token

1. Log in to your Confluence DC instance.
2. Click your avatar → Profile → Personal Access Tokens.
3. Click Create token. Name it (e.g. claude-mcp), set an expiry.
4. Copy the token immediately — it won't be shown again.

If the instance runs Confluence DC < 7.9, PATs are not available. Fall back to HTTP basic auth with CONFLUENCE_USERNAME + CONFLUENCE_API_TOKEN.

Available tools

Pages

Comments

Labels & Attachments

Users

confluence_get_page_views is Cloud-only. Do not call it on Data Center.

CQL (Confluence Query Language) reference

Used by confluence_search. Basic syntax:

field operator value [AND|OR field operator value ...]

Key fields on Data Center

Example CQL queries

type = page AND space = "ENG"
type = page AND text ~ "Kubernetes deployment"
type = page AND lastModified > now("-7d")
ancestor = 12345678 AND type = page
type = page AND space = "ENG" AND label = "architecture" AND lastModified > now("-30d")

DC-specific notes:

• Space keys are case-sensitive (usually uppercase).
• Personal spaces use ~username and must be quoted.
• content.property field is Cloud-only.

LATC Space — Pillar Page Structure

LATC Jira governance reference

Before proposing Jira-automation behavior for LATC workflows, cross-check against:

• Jira Usage at LATC (pageId=551394388)

This page defines current hierarchy semantics (including Feature usage), workflow expectations, and resolution guidance. Treat it as the source of truth when skill logic and legacy habits conflict.

Standard section layout

Pillar Homepage              ← pillar overview: vision, scope, deliverables, timeline, KPIs
├── 00-Regular Updates       ← status/progress updates, biweekly reports
├── 01-Resources & Working Model  ← team roster, headcount, working agreements, processes
├── 02-Architecture          ← architecture designs, system diagrams, design decisions
├── 03-Components            ← component-level specs, scopes, technical docs
└── 0x-...                   ← pillar-specific extensions

Known pillar homepages in LATC (page IDs)

Navigation pattern — use section-aware CQL

ancestor = <homepageId> AND title ~ "02-Architecture"
ancestor = <homepageId> AND title ~ "00-Regular Updates"

Known section page IDs for Operations & Infrastructure:

Known section page IDs for Data, Context & Memory:

Data Center vs Cloud key differences

Troubleshooting

Useful references

• mcp-atlassian source: https://github.com/sooperset/mcp-atlassian
• Confluence DC REST API: https://developer.atlassian.com/server/confluence/confluence-server-rest-api/
• Cursor MCP docs: https://docs.cursor.com/mcp