Skip to content

Chat

The chat interface is where you interact with your agents. It supports real-time streaming, tool use visualization, file uploads, and conversation management.

Open the chat by tapping the Chat button on an agent’s detail page, or by selecting an agent from the chat tab. The interface has three main areas:

  • Message list — Shows the conversation history with your messages on the right and agent responses on the left
  • Chat input — Text field at the bottom with send button and file attachment
  • Header — Agent name, new chat button, refresh history, and file panel toggle

Agent responses stream in real time using Server-Sent Events (SSE). As the agent generates text, it appears character by character in the message bubble. A typing indicator shows when the agent is processing.

You can stop a stream mid-response by tapping the stop button that appears while the agent is responding.

When the agent executes tools (like running bash commands, reading files, or fetching web pages), the chat shows tool use indicators inline. These show:

  • The tool name being executed
  • Tool input/parameters
  • Tool output/results

You can toggle tool call visibility using the Show Internals switch in the chat input bar, or from the Settings screen using the “Show Tool Calls” toggle.

On the web, you can attach files to your messages in two ways:

  • Click the attachment button in the chat input to open a file picker.
  • Drag and drop files directly onto the chat window.

Selected files appear as preview chips below the input field. Click the X on a chip to remove it before sending. When you send a message with files attached, they are uploaded to the agent’s workspace first, then the message is delivered. The agent can access and process these files using its file system tools.

Toggle the file panel from the chat header to browse files in the agent’s workspace. This gives you visibility into files the agent has created or modified.

  • New chat — Tap the new chat button in the header to clear the conversation history and start fresh. A confirmation dialog appears before clearing.
  • Refresh history — If the agent has been processing in the background (via subagents or heartbeat tasks), tap refresh to load any new messages.

When the agent spawns background subagents to handle complex tasks, the chat shows indicators for active subagent runs. Subagent results are delivered as pending messages that appear when you refresh or when they complete.

Each message consumes AI credits based on token usage. If you run out of credits, an error message appears with a prompt to upgrade your plan. Credit usage and limits are visible on the Settings/Account page.