inhouse_confluence_mcp

In-house Confluence (Data Center/Server) MCP server implementation. Default mode is read-only, and write tools are controlled by policy.

Tools

• confluence_list_spaces

• confluence_search_cql

• confluence_get_content

• confluence_get_labels

• confluence_get_children

• confluence_get_attachments

• confluence_get_comments

• confluence_scan_content

• confluence_get_likes (experimental, off by default)

• confluence_create_page (write, off by default)

• confluence_update_page (write, off by default)

• confluence_add_label (write, off by default)

• confluence_add_comment (write, off by default)

Related MCP server: Atlassian MCP Server

Quick Start

1. Copy .env.example to .env

2. Set required values:

• CONFLUENCE_BASE_URL

• CONFLUENCE_AUTH_MODE (pat or basic)

• CONFLUENCE_PAT or CONFLUENCE_USERNAME/CONFLUENCE_PASSWORD

3. Install and run:

python -m pip install -e .
confluence-mcp

Or:

python -m confluence_mcp.main

OpenCode Registration

Use OpenCode local MCP config with a command array and environment object.

{
  "mcp": {
    "inhouse-confluence": {
      "type": "local",
      "command": [
        "C:\\Users\\jae_chul.lee\\inhouse_confluence_mcp\\.venv\\Scripts\\python.exe",
        "-m",
        "confluence_mcp.main",
        "--env-file",
        "C:\\Users\\jae_chul.lee\\inhouse_confluence_mcp\\.env"
      ],
      "environment": {
        "PYTHONIOENCODING": "utf-8"
      }
    }
  }
}

Notes:

• Do not rely on Activate.ps1 in MCP commands.

• Prefer direct venv python path.

• --env-file makes startup independent from cwd.

• Optional override: set CONFLUENCE_MCP_ENV_FILE in environment.

.env Loading Rules

On startup, the server looks for .env in this order:

1. --env-file path

2. CONFLUENCE_MCP_ENV_FILE

3. Current working directory (.env, .env.local)

4. Python executable neighborhood (useful for .venv layout)

5. Project/module directory

If no file is found, startup logs searched paths.

Space Auto-Discovery

• If ALLOWED_SPACES is empty and AUTO_DISCOVER_SPACES=true, the server auto-loads available space keys on startup.

• DENIED_SPACES is optional. Use it only when you need an explicit sensitive-space blocklist.

• To generate a ready-to-copy env line manually:

confluence-mcp-sync-spaces

• Output file: logs/spaces_discovered.json

Recommended Settings

• Keep WRITE_ENABLED=false initially

• Configure ALLOWED_SPACES first

• Add DENIED_SPACES only for sensitive spaces

• Keep EXPERIMENTAL_LIKES=false unless needed

Logs

• Audit log path: AUDIT_LOG_PATH (default: logs/audit.jsonl)

• Fields: timestamp, tool, status, latencyMs, traceId, error

Tests

python -m unittest discover -s tests -v

Tool Probe on Startup

• By default, the server probes API permissions at startup and exposes only reachable tools.

• Toggle all probes with TOOL_PROBE_ON_STARTUP=true|false.

• Disable only space probe with SPACE_PROBE_ON_STARTUP=false.

• When SPACE_PROBE_ON_STARTUP=false, user-declared ALLOWED_SPACES is trusted for policy checks.

• If /rest/api/space is blocked, confluence_list_spaces is hidden.

• If /rest/api/content/search is blocked, read/write content tools are hidden.