USER MANUAL · OPERATE

Chat

Interact with your OpenClaw agents through a conversational interface — directly from ClawButler.

/chat

Overview#

The Chat page provides a conversational interface to interact with your OpenClaw agents without leaving ClawButler. Select an agent, type a message, and see the agent’s response in real-time — including any tool calls it makes along the way.

Chat is ideal for testing agent behavior, debugging tool call chains, and quick ad-hoc interactions. Messages are sent to the OpenClaw gateway via the connector tunnel and streamed back with full tool call visibility.

Prerequisites#

Prerequisites

At least one connector configured and synced (see Connectors)
Agents discovered and in a healthy state (see Agents)
Target agent must be running and enabled in OpenClaw

Expected Results#

After opening the Chat page and selecting an agent, you will see:

-Agent selector with health status indicators for each available agent
-Message thread area showing your messages and agent responses in chronological order
-Tool call expansion panels inline with the response, showing tool name, input, and output
-Input bar with send button at the bottom of the page

Web Operations#

Navigate to the Chat page from the sidebar under Operate.

Agent Selector

Use the dropdown at the top of the chat area to pick which agent to converse with. The selector shows agent name, emoji, and health status. Switching agents starts a new conversation thread.

Message Thread

Messages appear in a scrollable thread. Your messages are aligned to the right; agent responses appear on the left with the agent’s avatar. Responses are streamed in real-time so you can see the agent thinking. Markdown formatting in responses is rendered inline.

Tool Call Display

When the agent invokes a tool during its response, a collapsible panel appears inline showing the tool name, input parameters, and returned output. Expand any tool call to inspect the full payload. This is invaluable for debugging agent behavior and verifying correct tool usage.

CLI Operations#

Start an interactive chat session with the default agent

Terminal

$ ap chat start

Start a chat session with a specific agent by ID

Terminal

$ ap chat start --agent <agent-id>

Mobile Operations#

On mobile, go to the More tab to access Chat. Features include:

-Agent selector and full-screen message thread
-Real-time response streaming with tool call expansion
-Optimized keyboard handling for quick message input

FAQ#

Q: The agent is not responding to my messages.

A: Verify the agent is running and healthy in the Agents page. If the connector shows disconnected, reconnect it first. Also check that the agent is not disabled or killed via the kill switch.

Q: Can I see the token cost of a chat message?

A: Yes. Each agent response includes usage metadata. The cost is recorded in the Cost Tracking module and attributed to the agent. Check the Sessions page for per-message cost breakdowns.

Q: Do chat conversations persist across page reloads?

A: Chat sessions are maintained by the OpenClaw gateway. If you reload the page, the current session continues. However, navigating away and returning may start a fresh session depending on the agent’s session configuration.