Skip to content

How to Get Notifications from Your AI Agent

aimcptutorial

Your AI agent just finished a 30-minute task, but you didn't know until you checked back. Or worse, it needed your approval and sat idle for hours. Sound familiar?

BotBell solves this by giving your AI agent a direct line to your phone. Here are three ways to set it up, from easiest to most flexible.

Step 1: Create a Bot

Sign up at botbell.app, create a bot, and copy the Bot Token (bt_...). That's your agent's credential.

If your AI tool supports MCP, this is the simplest path. The agent can send messages and read your replies with zero code.

Add to your MCP config:

{
  "mcpServers": {
    "botbell": {
      "command": "npx",
      "args": ["-y", "@botbell/mcp-server"],
      "env": {
        "BOTBELL_TOKEN": "bt_your_token_here"
      }
    }
  }
}

The agent now has botbell_send and botbell_get_replies tools available automatically.

Option B: Agent Skill (One-line Install)

For Claude Code, GitHub Copilot, Cursor, Codex, Gemini CLI, and 30+ other AI tools:

npx skills add qq418716640/botbell-skill

The agent gets a /botbell command for notifications, confirmations, and questions. No MCP setup needed.

Option C: SDK (Full Programmatic Control)

For custom agents built with LangChain, AutoGPT, or your own framework:

from botbell import BotBell

bot = BotBell("bt_your_token_here")

# Send a notification
bot.send("Training complete! Accuracy: 94.2%")

# Ask a question and wait for reply
reply = bot.send_and_wait(
    "Deploy to production?",
    actions=[
        {"label": "Yes, deploy", "action": "deploy"},
        {"label": "No, cancel", "action": "cancel"}
    ]
)
print(reply.text)  # "deploy" or "cancel"

What's Next?

← Blog