# Using the chat: ask, attach, stop, and Ask agent How the agent's chat works: attaching files and screenshots, choosing a model, stopping a reply, typing while it answers, approvals, deleting or sharing a chat, and the Ask agent panel. **TL;DR:** Ask in the chat, or press ⌘I to open Ask agent beside the page you are on. Attach JPEG or PNG images and CSV files up to 4.5 MB, or a video. Stop ends a reply at its next step, and you pay only for the steps that finished. What you type while it answers waits in a queue and sends when the reply ends. URL: https://autron.ai/help/using-the-chat Updated: 2026-10-02 By: Autron --- Start a chat from the message box on Today or **New Chat** in the sidebar, or open **Ask agent** on the page you are looking at. It is the same agent, with the same history, approvals and credits, wherever you open it. ## Attach files, use skills, or dictate Press **+** in the message box: - **Add files or photos** attaches JPEG or PNG images and CSV files, each up to 4.5 MB. You can also paste a screenshot straight into the message box. - Videos are for Sponsored Brands video campaigns: MP4, MOV or M4V, up to 500 MB, 6 to 45 seconds, 16:9 or 9:16. [Video ads and creatives](https://autron.ai/help/video-ads-and-creatives) has the exact sizes. - **Skills** includes one of your saved skills with the message. Typing **/** opens the same list. Skills come with paid plans; see [Skills](https://autron.ai/help/skills). To dictate, press the microphone (or **Option+V**, **Alt+V** on Windows), speak, and press **Stop and transcribe**. Your words land in the message box for you to check, and nothing is sent until you send it. A recording can run for three minutes. ## Choose a model On a paid plan, the model picker sits next to **+**. It lists models from Anthropic, OpenAI, Google, xAI, DeepSeek and Z.ai, and each shows a cost tier from **$** to **$$$$**. More dollar signs mean each reply spends more credits. The app remembers your choice. The default is **GPT-6.1 Sol**, which did best on common seller requests in our testing, and it suits most work. A **$** model such as GPT-6 Luna costs less per reply for simple work. Save a **$$$$** model for a long, hard job where Sol falls short. The free plan has no picker. It answers on **GPT-6 Luna**, and once a conversation grows long it moves up to GPT-6.1 Sol by itself, which costs more per reply. Starting a new chat for a new topic keeps replies cheap. ## Stop a reply While the agent works, the send arrow turns into a square **Stop** button. Press it and the reply ends at its next step, or at once when no tool is running, and is marked **Stopped**. - You pay for the steps that finished before you pressed Stop. - A change already sent to Amazon cannot be recalled. The agent keeps its result in the chat, and the [Action log](https://autron.ai/help/action-log) records it. ## Type while it answers Anything you send while a reply is running waits in a queue above the message box. Queued messages send one at a time as each reply ends. Hover over one (on a phone they always show) for two icons: the up arrow, **Send now**, stops the reply and sends that message first, and the bin removes it. The agent still has everything the stopped reply had read. Pressing **Stop** holds whatever was queued at that moment. Those messages are marked **Held after Stop** and wait until you press their up arrow. Queued messages also wait while an approval is open. ## Paused, and Continue One reply can take up to 50 steps, within a time limit. At either limit the agent writes up what it has, and the reply ends with **Paused after 50 steps** or **Paused at the time limit for one reply**. Press **Continue** to carry on. It sends "Continue" as a new message. ## Approvals When the agent wants to make a change that your [Agent permissions](https://autron.ai/help/agent-permissions) do not already allow, it shows the approval card: **Allow once**, **Always allow**, **Block once** or **Always block**. A banner above the message box, **Approval pending**, repeats it with **Approve** and **Decline** in case you scrolled past. Sending a new message while an approval is waiting declines it. See [Confirmations and safety](https://autron.ai/help/confirmations-and-safety). ## Your chat history The sidebar lists your chats by date. Routine runs appear there too, marked with a bolt; the filter icon beside the first date heading has **Show routine runs** to hide them. A chat that started by email, in Slack or on WhatsApp carries a small mark for where it came from. Hover over a chat and its **⋯** menu has two actions: - **Delete** removes the chat for good, along with the files you uploaded to it. It cannot be undone. - **Share** sets the chat to **Private** or **Public**. > **Heads up** > > **Public means anyone with an Autron account.** Anyone signed in to Autron who has the link can read a Public chat, including people outside your company, along with the files in it. They cannot write in it. Set it back to **Private** to close it. ## Ask agent **Ask agent** opens the agent beside the page you are on, so you can ask about what you are looking at. On a computer it sits at the top right of every page except a full chat, Today included, and **⌘I** (**Ctrl+I** on Windows) opens it too. On a phone, tap **Ask agent** at the bottom of the screen and it opens as a drawer from the bottom. Today on a phone has no **Ask agent** button, because its own message box is right there. - It knows which page you are on and offers questions about it to start from. - **+** starts a new conversation. The current one stays in your history. - **⤢** is **Open as a full chat**, which moves the conversation to its own page once you have sent a message. - Closing the panel keeps the chat in your history. On a wide screen the panel sits beside the page and narrows it, and on a smaller one it slides over it. ## Credits Replies spend credits, though a quick lookup is free. The ring beside the send arrow shows how much you have left; hover over it for the detail. See [Credits and pricing](https://autron.ai/help/credits).