Bots
You now have an agent with instructions and tools — it can think, act, and get things done. The next question is: where do you talk to it?
The default: a web interface
Section titled “The default: a web interface”Most agent platforms give you a web interface where you can chat with your agents. This is where you configure and build them, and it is usually the first place you interact with them.
But it does not have to be the only place.
Bots: your agent, where you already are
Section titled “Bots: your agent, where you already are”The real power is making your agent available in the communication tools you already use — Telegram, Slack, WhatsApp, Microsoft Teams, or others.
This is the concept of a bot. You send a message to the bot on your messaging platform. The message is passed to your agent, which runs its loop on the server. When the agent has a response, it is delivered back to the same platform you used to send the message.
From your perspective, it feels like messaging a colleague — except this colleague has access to your tools, your data, and never forgets your instructions.
One agent, one bot
Section titled “One agent, one bot”There are different ways to set this up, but we recommend a clear mental model: each agent gets its own bot on each platform where you want it available.
Your task management agent might have a Telegram bot and a Slack bot. Your sleep coach agent might only live in the web interface. Your general assistant might be available everywhere.
This keeps things simple. When you message a bot, you know exactly which agent you are talking to and what it is capable of.
Some platforms support a single bot that routes messages to different agents behind the scenes. This can work, but it adds complexity. Start with one agent, one bot — you can always evolve from there.
Agent-to-agent communication
Section titled “Agent-to-agent communication”Bots are how you talk to your agents. But agents can also talk to each other.
An agent can be given the ability to ask another agent a question. For example, you message your general assistant via Telegram:
“How should I adjust my schedule today given how I slept?”
Your general assistant does not know about your sleep data. But it can ask your sleep coach agent, which fetches last night’s data, analyzes it, and sends a summary back. The general assistant then uses that information to give you a thoughtful answer.
From your side, this is a single conversation. Behind the scenes, two agents collaborated to help you.
This is powerful, but it is not something you need to set up right away. Start by making your agents available where you communicate. Agent-to-agent communication becomes valuable as your team of agents grows.
The key idea
Section titled “The key idea”Making your agents available as bots is what turns them from a tool you visit into a presence in your daily workflow. The easier it is to reach your agent, the more you will use it — and the more value it will provide.
In the next sections, we look at how to make your agents smarter and more autonomous — starting with skills and files.