Your First Conversation

With a model connected, you are ready to put the agent to work. A conversation in Wayland is more than chat: the assistant can read and write files, run commands, and search, with your approval.

Start a chat

From the dashboard:

1. Pick a starting point. You can type straight into the launch bar, choose an intent pill (Sell, Write, Research, Build, Run, Plan, Learn), or open a starter card.
2. Type a request and send it.

A good first task is something concrete and local, for example:

List the files in my Downloads folder and summarize what is there.

Approve tools as they run

When the assistant wants to use a tool, such as running a shell command or writing a file, it asks first. An inline approval prompt shows exactly what it intends to do. You can:

• Approve the single call,
• Approve and remember so similar calls run without prompting, or
• Deny and tell it to try another way.

This permission gate is on by default. You stay in control of anything that touches your system. To learn how approvals and auto-approve work, see the CLI flag reference (the same model applies in the desktop app).

Watch the work

As the assistant works, the conversation shows its messages, tool calls, and results. A workspace preview pane displays files it reads or edits so you can follow along without leaving the window.

Save and resume

Conversations are saved automatically. Find earlier ones in the Conversations list and pick up where you left off. The agent keeps the thread, including what it learned.

Where to go next

You now know the core loop: ask, approve, review. From here:

• Give a task its own persona with a custom assistant.
• Put several assistants on a job with teams.
• Connect more capabilities by installing MCP servers.