OpenCode

First complete Set up Jira and Confluence. Check your OpenCode version:

opencode --version

OpenCode 1.x and 2.x use different MCP configuration structures. Select the matching section below.

OpenCode 1.x

Add the server to your user configuration at ~/.config/opencode/opencode.json. On native Windows, this is under %USERPROFILE%\.config\opencode\opencode.json. You can also use a project-level opencode.json.

Merge this mcp entry with your existing settings. Replace YOUR_NAME and the executable path with your local values.

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "atlassian": {
      "type": "local",
      "command": [
        "/Users/YOUR_NAME/.local/bin/uvx",
        "mcp-atlassian==0.23.0",
        "--transport", "stdio",
        "--env-file", "/Users/YOUR_NAME/.config/tritonai/atlassian.env"
      ],
      "enabled": true
    }
  }
}

OpenCode puts the executable and all arguments in one command array. See the 1.x MCP reference and configuration guide.

OpenCode 2.x

The 2.x schema nests server definitions under mcp.servers and uses disabled. Use this form only for OpenCode 2.x:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "servers": {
      "atlassian": {
        "type": "local",
        "command": [
          "/Users/YOUR_NAME/.local/bin/uvx",
          "mcp-atlassian==0.23.0",
          "--transport", "stdio",
          "--env-file", "/Users/YOUR_NAME/.config/tritonai/atlassian.env"
        ],
        "disabled": false
      }
    }
  }
}

See the 2.x MCP reference and 2.x configuration guide.

Check the connection

Restart OpenCode and inspect the MCP server status. Then run the Jira and Confluence checks.

If OpenCode runs inside WSL, keep its configuration, uv installation, and environment file inside WSL. A desktop frontend connected to a WSL server uses that server's environment. See OpenCode's Windows and WSL guide.