Chat

Overview

The chat interface is the primary way to interact with CREAO. Type or speak a request in plain language, and the agent streams its response in real time — including text, code, generated files, and interactive artifacts.

Chat empty state showing the How can I help prompt with suggestion cards

The chat input

The input bar sits at the bottom of the screen. It includes:

Text input — describe what you need in natural language
Send button — click or press Enter to send
Stop button — appears during streaming; click to interrupt the agent
File upload — attach images, documents, and other files for the agent to process
Voice input — tap the microphone icon to dictate your message
@ mentions — type @ to reference skills and connectors by name

Close-up of the chat input bar showing the mic button, attachment button, and send button

Voice input

Tap the microphone icon in the input bar to dictate your message instead of typing. The input bar transforms into a recording interface with a live waveform visualization, a timer, and confirm/cancel buttons.

Voice recording UI showing waveform bars, a 0:00 timer, cancel (X) and confirm (checkmark) buttons

Start recording

Tap the microphone icon in the input bar. Your browser will ask for microphone permission the first time. A brief “activating” spinner appears while the microphone connects.

Speak your message

A waveform visualization appears with a timer showing how long you’ve been recording. Speak naturally — the recording supports up to 2 minutes.

Stop and transcribe

Tap the checkmark to stop recording and send the audio for transcription. A “Transcribing…” indicator appears while the audio is processed.

Review and send

The transcribed text appears in the input field. You can edit it before sending, or tap the mic again to append more dictation.

You can also tap the X button to cancel the recording and discard the audio without transcribing.

Voice input uses OpenAI’s speech-to-text model for high-accuracy transcription. It supports multiple languages and works in all major browsers.

Sending messages

Type your request and press Enter. The agent begins streaming its response immediately.

Build a landing page for a SaaS product with a hero section, pricing cards, and a contact form

Be specific about what you want. Include details like file format, styling preferences, and functional requirements.

Streaming responses

Responses stream token by token in real time. During streaming you’ll see:

Thinking indicator — animated dots while the agent plans its approach
Status updates — messages like “Starting sandbox…” or “Running agent…”
Tool execution — each tool call and its output stream in real time, so you can see exactly what the agent is doing as it works
Text content — the response appears progressively
Artifact panel — generated files open in an interactive preview on the right

Chat showing a user message, web search tool execution with results, and streamed text response

Interrupting the agent

Click the Stop button (red square) at any time to cancel the current response. The agent stops immediately and you can send a new message.

Queued follow-up messages

You can send additional messages while the agent is still processing your previous request. Instead of waiting for the agent to finish, type your follow-up and send it — it enters a queue and is processed automatically after the current turn completes. Queued messages support two modes:

Automatic — the agent picks up the next queued message and processes it server-side. You can close your browser and the queue continues.
Manual — the next queued message is promoted into the input box for you to review and edit before sending.

This is useful for multi-step workflows where you already know the next few instructions. Queue them all up front and let the agent work through them in sequence.

Extended thinking

When using a thinking-capable model (such as Claude Sonnet 4.6 or Claude Opus 4.6), you can enable Max Thinking to give the agent more time to reason through complex problems before responding. This is especially useful for multi-step tasks, intricate code generation, and nuanced analysis. To enable it, click the model name in the bottom-left corner of the chat input to open the model selector. The Max Thinking toggle appears at the top of the dropdown when a supported model is selected.

Model selector dropdown showing the Max Thinking toggle with a brain icon, followed by available models from Anthropic, Google, and OpenAI

Max Thinking uses more credits per response. If the model runs out of thinking budget mid-response, the agent automatically retries to ensure you get a complete answer.

Artifacts in chat

When the agent generates an HTML page, SVG, PDF, or other file, an artifact panel opens on the right side of the screen with an interactive preview. An artifact chip also appears in the chat message so you can reopen the panel later.

Split view with chat on the left and an interactive HTML artifact preview on the right

Generated HTML pages are fully interactive — buttons, forms, modals, and localStorage all work in the preview panel. You can also switch to Code view, download the file, or copy the source.

File uploads

Click the paperclip icon or drag files into the chat to attach them. Supported file types include:

Images (PNG, JPG, GIF, WebP)
Documents (PDF, TXT, CSV)
Code files (HTML, CSS, JS, Python, etc.)

The agent can read, analyze, and transform uploaded files as part of its response.

@ Mentions

Type @ in the chat input to open a menu of available skills and connectors. Select one to reference it in your message:

@web-search Find the latest React best practices for form handling

This tells the agent to use a specific skill for the task.

Text-to-speech playback

Click the three-dot menu (…) on any assistant message and select Summarize & read aloud. The agent automatically summarizes long or complex responses into concise spoken sentences and plays them back as natural-sounding audio. Click again to stop playback.

Message action menu showing Cost and Summarize and read aloud option with speaker icon

Share any conversation as a read-only public link. Click the share icon in the thread header to generate a shareable URL. Anyone with the link can view the full conversation — including messages, artifacts, and generated files — without needing a CREAO account. Shared threads are read-only. Recipients can browse the conversation and download artifacts, but cannot send messages or modify the thread. To revoke access, click the share icon again and disable the link. Previously shared URLs will stop working immediately.

Conversation history

Each workspace maintains its own conversation history. You can:

Continue a conversation — click any thread in the sidebar to resume it
Start fresh — click the New chat button to begin a new thread
Pin important threads — keep frequently referenced conversations accessible

Thread sidebar

The left sidebar shows your recent threads organized by workspace. Each thread displays its title (auto-generated from the first message) and timestamp.

Workspaces

Workspaces are the organizational unit for your conversations, skills, connectors, and agents. Switch between workspaces using the dropdown at the top of the sidebar. Each workspace has its own:

Conversation history
Enabled skills and connectors
Agents
File storage

You must select a workspace before using the chat. The agent’s available capabilities depend on which skills and connectors are enabled in the active workspace.

Language support

The chat interface is available in multiple languages including English, Portuguese (Brazil), Chinese (Simplified), French, Spanish, and Japanese. Your browser language is detected automatically, and the entire UI — including buttons, menus, status messages, and labels — adapts to your preferred language.

Keyboard shortcuts

Shortcut	Action
Enter	Send message
Shift + Enter	New line in message
@	Open skill/connector mention menu

Get Started

Features

Agent Brain

Pro

Community and Updates

Overview

The chat input

Voice input

Sending messages