I'm heading on a trip soon. Not the kind with airport lounges and hotel WiFi. The kind where connectivity is spotty at best. A bar of signal here, a tethered phone there.

That got me thinking. My agent is useful precisely because it can look things up, send emails, check my calendar. But all of that lives behind a web app. If I can barely load a page, I can't use it.

Then I remembered that Telegram is built for exactly this. Messages queue up and send when they can. Voice notes are tiny files. Photos compress automatically. The whole app is designed to work on bad connections.

So I connected my agent to Telegram. Now it lives in my chat list, alongside friends and family. No app to switch to, no page to load. Just a conversation.

Setting it up

In The Agency, go to Settings and tap Connect Telegram. You'll get a six-digit code. Open Telegram, find the bot, and send the code. That's the entire setup, about 30 seconds.

From that point on, your agent is reachable in Telegram. Same agent, same memory, same tools. Everything it does through Telegram shows up in your web dashboard too, so nothing is lost.

Text: the basics

The simplest interaction is a text message. I can ask my agent to check my calendar, draft an email, look something up. The same things I'd do in the web app. It processes the request, uses whatever tools it needs, and sends back a response.

The experience is surprisingly natural. Conversation context carries over, so follow-ups work. It feels less like using a tool and more like texting someone who happens to have access to all your stuff.

Voice: talk when you can't type

This is where it gets interesting for travel. Telegram has built-in voice messages. Hold a button, speak, release. Your agent receives the audio, understands what you said, and responds.

Imagine standing in a market somewhere and asking your agent:

How do you say "can I have the bill, please" in Japanese?

Just hold the button, say it, let go. A few seconds later, the answer is in your chat. No typing, no switching apps, no keyboard fumbling in bright sunlight.

Voice messages are tiny. A 30-second message is barely a few kilobytes. They send even on the weakest connections.

Photos: point and understand

This is the one I'm most looking forward to on this trip. Snap a photo of a menu, a street sign, a train schedule, a museum plaque. Send it to your agent and ask what it means.

Your agent sees the image, reads the text, understands the context, and explains it in plain language. One photo, one message, one answer.

No juggling a camera app, a translation app, and a browser. No copying text from a photo into a translator and hoping it gets the nuance right. Your agent has the full picture, literally, and can explain it the way a local friend would.

What I'm actually planning to use it for

On this trip, my phone will be my primary device. Here's what I expect to lean on:

Menus and signs — photo of a restaurant menu, "what are the vegetarian options?" Or a photo of a transit map, "which line gets me to the old town?"
Quick translations — voice message asking how to say something, getting the phrase back in text I can show someone
Staying on top of things back home — "any important emails today?" or "reschedule my Thursday meeting to next week"
Travel logistics — "what's the weather like tomorrow?" or "find me a train from X to Y in the morning." Connect the Google Maps app and your agent can send you real-time directions and navigation links directly in the chat.

All through the same lightweight chat interface, all on minimal bandwidth.

It's still the same agent

The key thing is that this isn't a separate bot or a limited version. It's the same agent, same memory, same connected apps, same personality. Tell it something in Telegram and it still knows that when you open the web app on your laptop.

Each channel keeps its own conversation thread, so a Telegram chat stays in Telegram. What travels with you is everything your agent has learned.