Let your agent talk to any human

Caspian gives your AI agent one identity that reaches people wherever they already are (email, Slack, Discord, WhatsApp, SMS, X, Telegram, iMessage), all behind a single on_message handler. You write the handler once; Caspian handles every provider's quirks, threading, delivery, and dedup.

The one-handler model

Every channel (a Slack thread, an SMS, a WhatsApp chat, an email) arrives at the same handler. You reply with message.reply(), and Caspian sends it back on the right channel, in the right thread, formatted the way that platform expects.

from caspian_sdk import CommClient

client = CommClient()               # reads COMM_API_KEY from env / .env

client.connect_email()              # one line per channel, no handler changes

@client.on_message
def handle(message):
    message.reply(f"You said: {message.text}")

client.listen()                     # one loop, every channel

import { CommClient } from "caspian-sdk";

const client = new CommClient();           // reads COMM_API_KEY from env / .env

await client.connectEmail();               // one line per channel

client.onMessage(async (message) => {
  await message.reply(`You said: ${message.text}`);
});

await client.listen();                     // one loop, every channel

Adding a channel is another connect_*() call, never new handler code.

Channels

Free channels cost nothing and need no signup. Paid channels run on Caspian's own network and unlock after a one-time sign-in.

How it fits together

• Connections: each channel you connect (an inbox, a Slack app, a bot) is a connection under your project.
• Conversations: inbound messages group into conversations with a stable id, so your agent can key its memory on a thread.
• One handler: on_message fires for every connection; reply() routes back to the source channel automatically.
• Behavior guides: Caspian can hand your agent per-channel etiquette (Slack threads, WhatsApp's 24-hour window, SMS length) to inject into its system prompt.

Next steps