Your First Conversation

You’ve installed an agent and sent your first message. Now let’s dig a little deeper into what’s actually happening when you have a conversation — and how to get the most out of it.

Starting a conversation

Open your agent from the Home screen. Every time you click into an agent, you start a new conversation thread. Previous conversations are listed in the sidebar so you can always go back.

Try asking your agent something that requires a bit of thought:

“What’s the weather like this week in Austin, Texas?”

Depending on your agent’s capabilities, one of two things will happen:

1. If the agent has a weather capability enabled, it will use that tool to fetch real data and give you an actual answer.
2. If it doesn’t have that capability, it will let you know it can’t look up live weather and offer to help in other ways.

This is an important idea in Selu: agents are only as powerful as the capabilities you give them.

How capabilities work

Capabilities are the tools your agent can use to take actions beyond just generating text. Think of them as skills you grant to your agent.

What are capabilities?

A capability is a specific tool — like web search, file reading, calendar access, or API calls. Each one is a small gRPC service that your agent can invoke during a conversation.

You're in control

You decide which capabilities each agent has. An agent can only use tools you’ve explicitly enabled. Nothing runs without your permission.

When your agent decides it needs to use a capability, here’s what happens:

1. The agent receives your message and thinks about how to respond.
2. It determines that answering well requires using a tool (for example, searching the web).
3. It sends a structured request to the capability service via gRPC.
4. The capability executes the action and returns the result.
5. The agent incorporates that result into its response to you.

All of this happens in a few seconds. You’ll see the agent’s response appear in the chat — sometimes with a small indicator showing which capability was used.

Watching it happen in real time

Selu includes a conversation inspector that lets you peek behind the curtain. To open it:

1. While in a conversation, click the Inspector icon in the top-right corner of the chat window.
2. You’ll see a panel showing each step the agent took — the messages exchanged, any capability calls, and the raw model responses.

This is especially useful when you’re curious about why an agent responded a certain way, or when you’re debugging a custom agent.

Multi-turn conversations

Agents in Selu maintain context within a conversation. This means you can have natural, multi-turn exchanges:

You: “Can you help me draft an email to my team about the project deadline?”

Agent: “Sure! What’s the deadline, and what tone are you going for — formal or casual?”

You: “Friday the 15th. Keep it casual.”

Agent: “Here’s a draft: …”

The agent remembers what you said earlier in the thread. If you start a new conversation, it starts fresh with no memory of previous threads.

Note

Conversation history is stored locally on your Selu instance. It’s never sent to external services unless a specific capability requires it (and you’ll see that in the inspector).

Tips for better conversations

• Be specific. Instead of “help me with my project”, try “help me outline the next three milestones for my website redesign.”
• Check capabilities. If an agent can’t do something you expect, go to its settings and see what capabilities are enabled. You might need to install or enable one.
• Use the right agent for the job. A general assistant is great for everyday tasks, but a specialized agent (from the marketplace) will often do better for domain-specific work.
• Iterate. Agents respond well to follow-up. If the first answer isn’t quite right, refine your request rather than starting over.

Conversations across channels

One of the most useful features of Selu is multi-channel messaging. If you’ve connected Telegram or iMessage in your settings, you can chat with the same agent from your phone.

Web chat

The built-in dashboard chat. Best for longer sessions where you want access to the conversation inspector and agent settings.

iMessage

Text your agent just like you’d text a friend. Great for quick questions on the go.

Telegram

Use Telegram’s bot interface to chat with your agents. Useful if Telegram is already part of your workflow.

Conversations started on different channels are separate threads, but they all connect to the same agent with the same capabilities.

Next steps

Now that you understand how conversations and capabilities work, you might want to:

• Explore the marketplace to find agents tailored to your needs
• Learn about capabilities in the Developer Guide to build custom tools
• Connect more channels so you can reach your agents from anywhere