Jira Confluence Atlassian Interactor

Standalone MCP server for Jira and Confluence.

Use JIRA_PAT and CONFLUENCE_PAT for Jira/Confluence Data Center/Server (Bearer token), or ATLASSIAN_PAT as a shared fallback for both products. For Cloud, use ATLASSIAN_USERNAME + ATLASSIAN_API_TOKEN.

Requirements

• Node.js 20 or newer
• Jira and/or Confluence
• A PAT created in Atlassian for Data Center/Server, or an Atlassian Cloud API token with username

Quick Start

1. Create either a Data Center/Server PAT or, for Cloud, use your Atlassian username and API token.
2. Add the server to your MCP client using stdio.
3. Put the PAT values in the MCP client's local environment configuration, a local shell environment variable, or a secret manager if your client supports one.
4. Run atlassian_validate_connection from your MCP client.

Never commit a PAT to git. Avoid putting it in any tracked config file, issue comment, Confluence page, bug report, or log.

MCP Client Configuration

This server runs as a local stdio MCP server. All clients need the same command and environment variables:

{
  "mcpServers": {
    "atlassian-dc": {
      "command": "npx",
      "args": ["-y", "jira-confluence-atlassian-interactor"],
      "env": {
        "JIRA_BASE_URL": "https://jira.example.company.com",
        "CONFLUENCE_BASE_URL": "https://confluence.example.company.com",
        "JIRA_PAT": "${JIRA_PAT}",
        "CONFLUENCE_PAT": "${CONFLUENCE_PAT}"
      }
    }
  }
}

For local-only client configuration, the env block is the most direct place to put product URLs and PAT values. Prefer storing the PAT values in local environment variables or a secret manager when your MCP client supports it.

If both products can share the same bearer token, you can set ATLASSIAN_PAT instead of JIRA_PAT and CONFLUENCE_PAT. Product-specific variables take precedence when both are present.

Configure at least one product URL. Jira tools require JIRA_BASE_URL; Confluence tools require CONFLUENCE_BASE_URL.

Client Setup Examples

Use the examples below as starting points. Replace the URLs with your Data Center/Server instances. For PAT values, prefer ${JIRA_PAT} / ${CONFLUENCE_PAT} / ${ATLASSIAN_PAT} or your client's secret input support over pasting tokens into a repository file.

OpenAI Codex

Codex reads MCP servers from ~/.codex/config.toml or a project .codex/config.toml.

[mcp_servers.atlassian-dc]
command = "npx"
args = ["-y", "jira-confluence-atlassian-interactor"]
enabled = true

[mcp_servers.atlassian-dc.env]
JIRA_BASE_URL = "https://jira.example.company.com"
CONFLUENCE_BASE_URL = "https://confluence.example.company.com"
JIRA_PAT = "${JIRA_PAT}"
CONFLUENCE_PAT = "${CONFLUENCE_PAT}"

You can also manage servers with the Codex CLI:

codex mcp add atlassian-dc -- npx -y jira-confluence-atlassian-interactor

If you use the CLI command, add the required environment variables in your Codex config afterward.

Claude Code

Claude Code can add stdio MCP servers from the terminal. Put options before the server name, then put the server command after --.

export JIRA_PAT="<YOUR_JIRA_PAT>"
export CONFLUENCE_PAT="<YOUR_CONFLUENCE_PAT>"

claude mcp add --transport stdio --scope user \
  --env JIRA_BASE_URL=https://jira.example.company.com \
  --env CONFLUENCE_BASE_URL=https://confluence.example.company.com \
  --env JIRA_PAT="$JIRA_PAT" \
  --env CONFLUENCE_PAT="$CONFLUENCE_PAT" \
  atlassian-dc -- npx -y jira-confluence-atlassian-interactor

Verify inside Claude Code with:

/mcp

Claude Desktop

Open Claude Desktop's MCP configuration from the app settings, then add a local stdio server entry:

{
  "mcpServers": {
    "atlassian-dc": {
      "command": "npx",
      "args": ["-y", "jira-confluence-atlassian-interactor"],
      "env": {
        "JIRA_BASE_URL": "https://jira.example.company.com",
        "CONFLUENCE_BASE_URL": "https://confluence.example.company.com",
        "ATLASSIAN_PAT": "${ATLASSIAN_PAT}"
      }
    }
  }
}

Restart Claude Desktop after editing the config.

Cursor

Open Cursor Settings, go to MCP, and add a new server. For JSON-based setup, use:

{
  "mcpServers": {
    "atlassian-dc": {
      "command": "npx",
      "args": ["-y", "jira-confluence-atlassian-interactor"],
      "env": {
        "JIRA_BASE_URL": "https://jira.example.company.com",
        "CONFLUENCE_BASE_URL": "https://confluence.example.company.com",
        "ATLASSIAN_PAT": "${ATLASSIAN_PAT}"
      }
    }
  }
}

Keep PAT-bearing Cursor config local. Do not commit workspace MCP files that contain real tokens.

OpenCode

OpenCode reads JSON/JSONC config from ~/.config/opencode/opencode.json (or project opencode.json / .opencode/opencode.json). Add local MCP servers under mcp with type: "local". This server accepts JIRA_PAT and CONFLUENCE_PAT, with ATLASSIAN_PAT as a shared fallback.

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "atlassian-dc": {
      "type": "local",
      "command": ["npx", "-y", "jira-confluence-atlassian-interactor"],
      "enabled": true,
      "environment": {
        "JIRA_BASE_URL": "https://jira.example.company.com",
        "CONFLUENCE_BASE_URL": "https://confluence.example.company.com",
        "JIRA_PAT": "{env:JIRA_PAT}",
        "CONFLUENCE_PAT": "{env:CONFLUENCE_PAT}"
      }
    }
  }
}

Set JIRA_PAT and CONFLUENCE_PAT in your shell before launching OpenCode, or use ATLASSIAN_PAT when both products share the same token, then run:

opencode mcp list

VS Code / GitHub Copilot Agent Mode

VS Code uses an mcp.json file with a top-level servers object. For workspace config, create .vscode/mcp.json; for user config, run MCP: Open User Configuration.

{
  "servers": {
    "atlassian-dc": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "jira-confluence-atlassian-interactor"],
      "env": {
        "JIRA_BASE_URL": "https://jira.example.company.com",
        "CONFLUENCE_BASE_URL": "https://confluence.example.company.com",
        "ATLASSIAN_PAT": "${ATLASSIAN_PAT}"
      }
    }
  }
}

VS Code can also add a server from the command line:

code --add-mcp '{"name":"atlassian-dc","type":"stdio","command":"npx","args":["-y","jira-confluence-atlassian-interactor"]}'

Add the environment variables in the generated config afterward.

Windsurf / Cascade

Windsurf Cascade reads local MCP servers from ~/.codeium/windsurf/mcp_config.json.

{
  "mcpServers": {
    "atlassian-dc": {
      "command": "npx",
      "args": ["-y", "jira-confluence-atlassian-interactor"],
      "env": {
        "JIRA_BASE_URL": "https://jira.example.company.com",
        "CONFLUENCE_BASE_URL": "https://confluence.example.company.com",
        "ATLASSIAN_PAT": "${ATLASSIAN_PAT}"
      }
    }
  }
}

Set ATLASSIAN_PAT in your shell or OS environment before launching Windsurf.

Setup Notes

• Use a PAT from your own Atlassian user profile.
• If Jira and Confluence use different hosts, set both base URLs.
• If only one product is needed, omit the other base URL.
• Restart the client after editing MCP config unless the client has an explicit MCP restart command.
• Run atlassian_validate_connection first; it should return authenticated user details for each configured product.
• Never commit real PATs. Use user-level config, prompt inputs, environment variables, or a secret manager.

PAT handling differs by client: ${ATLASSIAN_PAT} where supported, while some clients may require prompt-based input blocks such as ${input:...}.

Reference Docs

• Codex MCP configuration
• Claude Code MCP
• Cursor MCP
• OpenCode MCP servers
• VS Code MCP servers
• VS Code MCP configuration reference
• Windsurf Cascade MCP

Verify Setup

Run the atlassian_validate_connection tool from your MCP client. It should return authenticated Jira and/or Confluence user details.

Documentation

• docs/config.md: environment variables
• docs/security.md: PAT handling
• docs/tools.md: tool list and write safety