Start your first conversation
You've installed MultiClaw and connected your workspace. Now it's time to talk to your first agent. Open the app, pick an agent, and type your first message.
Before you start
Make sure two things are in place:
- Active connection: Check the gateway status indicator in the top-right corner. It should be green. If it shows red, connect to your workspace first.
- At least one agent: If you haven't created one yet, see Create an agent.
Start a conversation
When no conversation is active, you'll see a welcome greeting — "What can I help you build today?" — with the composer centered below it.
- In the composer toolbar, click the agent name to open a searchable list, then select the agent you want to use.
- Type your message in the composer.
- Press Enter to send.
After you send your first message, the composer moves to the bottom of the screen. The agent's reply streams in real time, word by word. When the stream finishes, type a follow-up message to continue the conversation.
Not sure what to type? Start with something specific: "Create a Python script that lists all files in a folder" or "Draft an email introducing our new product." The more context you give, the better the response.
When you send a message, its content is processed by the AI model provider you have configured (such as OpenAI, Anthropic, or Google). Those providers handle this data under their own privacy policies. For details on how MultiClaw handles your data, see the MultiClaw Privacy Policy.
You can have multiple conversations with the same agent. Each conversation is a separate thread with its own history.
Start a new conversation
To open a fresh conversation at any time:
- Click the pencil icon in the top-left corner of the chat area, or
- Press Cmd+N (Mac) / Ctrl+N (Windows/Linux).
The agent selector locks after you send the first message. You'll see a lock icon next to the agent name. To use a different agent, start a new conversation.
When the agent shows a plan
If your request requires action — using the browser, editing files, or sending a message — the agent drafts a plan before doing anything. The plan appears in the conversation as a list of steps the agent intends to take.
Review the steps, then approve the plan to let the agent proceed. The agent does not proceed until you approve.
For simple questions — like "What does this error mean?" — no plan appears. The agent answers directly in the conversation.
Keyboard shortcuts
| Action | Shortcut |
|---|---|
| Send message | Enter |
| New line in message | Shift+Enter |
| New conversation | Cmd+N (Mac) / Ctrl+N (Windows/Linux) |
Related articles
Explore the interface
A mental map of the MultiClaw window: navigation sidebar, title bar, main content, conversations, Chats sidebar, and composer.
Quick chat vs tasks
Understand when the agent replies conversationally and when it takes action on your behalf.
What are agents?
Agents are AI assistants that plan, act, and complete tasks on your behalf — here's what they can do and how they work.