MCP server: manage your Discord server from AI assistants

Connect a client

The server URL is https://console.botgg.app/mcp (MCP over Streamable HTTP). Clients sign in in one of two ways:

• Sign in with Discord (OAuth) — recommended. The client only needs the URL. It opens BotGG in your browser, you log in with Discord, pick one server and the permissions, and approve. Nothing to copy, and you can disconnect the app at any time.
• API key. For clients without OAuth support: create a key in the console (see authentication) and send it as Authorization: Bearer botgg_….

Claude (claude.ai and Claude Desktop)

1. Open Settings → Connectors and choose Add custom connector.
2. Name it BotGG and paste https://console.botgg.app/mcp as the URL, then Add.
3. Click Connect: BotGG opens, you log in with Discord and approve (see the consent screen).

ChatGPT

1. Open Settings → Apps & Connectors. Custom connectors need Developer mode (under Advanced settings) on your plan.
2. Choose Create, name it BotGG, paste https://console.botgg.app/mcp as the MCP server URL and pick OAuth as the authentication.
3. Create the connector and sign in when ChatGPT opens BotGG.

Cursor

Add the server to ~/.cursor/mcp.json (or the project’s .cursor/mcp.json). Cursor shows Needs login next to it — click it to sign in with Discord in the browser.

{
  "mcpServers": {
    "botgg": { "url": "https://console.botgg.app/mcp" }
  }
}

Claude Code

claude mcp add --transport http botgg https://console.botgg.app/mcp
# then run /mcp inside Claude Code and choose "Authenticate"

Any client with an API key

{
  "mcpServers": {
    "botgg": {
      "type": "http",
      "url": "https://console.botgg.app/mcp",
      "headers": { "Authorization": "Bearer botgg_…" }
    }
  }
}

The consent screen

When a client signs in, BotGG shows who is asking and where you will be sent back afterwards. Apps register themselves, so their name is not verified — only approve a connection you started yourself.

1. Log in with Discord (if you aren’t logged in to the console already).
2. Choose one server. Only servers where you have Manage Server are listed, and BotGG must be added to the one you pick. A connection always belongs to exactly one server.
3. Choose permissions. They are the same scopes as for API keys. Sensitive ones (acting on members, moderation data) start unchecked; selecting one needs an extra confirmation.
4. Approve or deny. The app receives a short-lived access token (1 hour) and a refresh token that changes on every use.

Every connection is listed in the console under Server → Connected apps with its permissions, who connected it and when it was last used. Any manager of the server can disconnect it there; the app loses access immediately. If the person who connected the app loses Manage Server, its access ends automatically on the next call.

Tools

There is one tool per REST API endpoint — for example bank.getBalance becomes the tool bank_get_balance. Inputs are the endpoint’s path, query and body fields; the server ID is never an input, it comes from the connection. The tool list always matches the API reference, and a client only sees the tools whose scope its connection has.

• Read-only tools are marked as such; tools that change or remove data are marked destructive, and the assistant is told to confirm the exact action with you first.
• Every call runs through the same pipeline as the REST API: scope and feature checks, validation, error codes and the plan’s rate limit for the connection (an API key shares one budget between REST and MCP).
• Errors come back as tool results with the same error codes as the REST API, so the assistant can explain or fix them.
• Changes made through MCP are written to the server’s audit log with the app’s name.

Limits

Safety

• Confirmation of destructive actions is done by your AI app, based on the tool markings — BotGG cannot see your chat. Keep confirmations turned on in your client.
• Write calls over MCP don’t use an Idempotency-Key, so an assistant shouldn’t blindly repeat a failed write — check the result first. For automated, retried writes use the REST API.
• Tokens are bound to this server URL and to one Discord server; BotGG stores only their hashes. A reused refresh token disconnects the app as a precaution — just connect it again.

Troubleshooting

HTTP/1.1 401 Unauthorized
WWW-Authenticate: Bearer resource_metadata="https://console.botgg.app/.well-known/oauth-protected-resource/mcp"

HTTP/1.1 403 Forbidden
WWW-Authenticate: Bearer resource_metadata="…", error="insufficient_scope", scope="bank:write"