Connect OpenCode to XgenPlus Mail

When to use it

Arjun writes code in OpenCode at acme.in. He asks it to find this week's bug reports from customers and turn them into a task list in his project.

Configuration

[Convert]::ToBase64String([Text.Encoding]::UTF8.GetBytes('you@xgenplus.com:AGENT_PASSWORD'))

printf '%s' 'you@xgenplus.com:AGENT_PASSWORD' | base64 | tr -d '\n'

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "xgenplus": {
      "type": "remote",
      "url": "https://mcp.xgenplus.com/mcp",
      "enabled": true,
      "headers": { "Authorization": "Bearer {env:XGENPLUS_TOKEN}" }
    }
  }
}

Keep {env:XGENPLUS_TOKEN} exactly as shown. OpenCode fills in the token from the environment variable when it starts.

Good to know

• OpenCode writes environment variables as {env:NAME}, not ${env:NAME} as Cursor does. With the wrong form, or with the variable not set, OpenCode sends an empty token and the server refuses the connection.
• When the server refuses a token, OpenCode may offer an OAuth sign-in. XgenPlus does not offer OAuth sign-in yet, so check the token and the variable instead.
• "max_tokens cannot exceed ..." errors come from your AI model's provider, not from XgenPlus. OpenCode asked for a longer answer than the model allows. Set a lower output limit for that model in opencode.json, below the provider's limit for that model:

"provider": {
  "<provider>": { "models": { "<model>": { "limit": { "context": 128000, "output": 8000 } } } }
}