> ## Documentation Index
> Fetch the complete documentation index at: https://help.get-ryze.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Chat with the AI Marketer
> Open the agent from any page, attach files, choose how much it may change without asking, and find past chats and agent tasks in Chat History.
The chat is where you talk to the **AI Marketer** — ask a question, give it a task, or follow up on its work. It reads your connected accounts and can make changes in them, so it's worth knowing how the chat works.
## Two ways to open it
* **The AI Marketer button** in the top bar (or **⌘J** / **Ctrl+J**) opens the agent in a side panel on whatever page you're on. In the panel you'll find **Recent chats (last 20 chats)**, **New chat**, and **Open as full chat** to move the conversation to a full page.
* **AI Marketer → New chat** in the sidebar opens a full-page chat.
Many buttons across the app open the side panel and send their prompt to the agent right away — for example **Create with Agent**, **Generate Report**, **Update Report**, **Improve Brand Profile**, **Fix with agent** and any card on the [Templates](/agent/templates) page. That run starts immediately and uses credits like any other message (see [Credits & Usage](/agent/credits)).
## Write a message
* **Attach files.** In a full-page chat, click **+** → **Attach files** → **Upload from computer**, **Choose from Media** or **Upload from Drive**. In the side panel, the paperclip button opens the same three options. You can also drag files onto the chat.
* **Reference ads** (in the **+** menu) attaches your own ads or ad templates as references for a new creative. **Image aspect** sets the shape of the creatives the agent makes, or leaves it on **Auto**.
* **Record voice** — click the microphone, speak, then **Stop & send**. Your recording is transcribed into the message box.
* **Stop generating** — the square button stops the reply while the agent is still answering.
Attachments can be images, PDF, DOCX, XLSX or CSV files — up to 3 images and 3 documents per message. Images can be up to 10 MB each, documents up to 20 MB. Images from Google Drive aren't supported — upload them from your computer instead.
## Tool approvals
The **Tool approvals** picker in the message box decides how much the agent may do in your connected accounts without asking:
| Option | What happens |
| - | - |
| **Manual approve** | Every tool that touches your data needs your approval. |
| **Approve write & delete** | Reads run freely; changes need your approval. |
| **Skip all approvals** | The agent runs every tool on its own. |
The default is **Skip all approvals**, so the agent can change things in your connected accounts — budgets, campaigns, content — without asking first. If you want to confirm changes, switch to **Approve write & delete** or **Manual approve**. Your choice is remembered in this browser.
When a tool is waiting for you, the chat shows **Approve** and **Deny** (or **Approve all** and **Deny all** when there are several).
## Agent tasks
Some actions, such as **Fix with agent** on a Technical Audit item, start an **agent task**: the agent works through the job on its own and reports back when it's done, and you can leave while it runs. **Stop agent** halts it after its current step. Agent tasks are billed like any agent run and are listed under **Chat History → Agent tasks**.
## While you chat
* **Assets** — the panel at the top right of a chat has three sections: **Tasks** (the agent's to-do list for this chat), **Generations** (creatives and reports made in this chat) and **Context** (files you attached).
* Under each reply: **Copy**, and the coin icon showing **N credits used**. Hover it for the breakdown.
* **Chat options** menu at the top: **Rename** or **Delete**. Deleting removes the chat and its messages for good.
* If you see "Long chats slow the agent down — a fresh one keeps it sharp.", click **New chat** and continue there.
The agent can make mistakes. Review what it produces before you rely on it.
## If something goes wrong
| Message | What to do |
| - | - |
| **Out of credits** | Click **Buy credits**, or **Add your key** to run chat on your own [Anthropic key](/integrations/anthropic). |
| **Workspace credit cap reached** | This workspace used its monthly credit limit. The organization owner can raise it in **Organization Settings → Usage**. |
| **Paid Ads is inactive** | Your subscription is inactive. Click **Manage billing**. |
| **Conversation is too large** | Remove large attachments or start a new chat. |
| **Agent is over capacity** | The AI provider is briefly overloaded. Retry in a few seconds. |
| **Something went wrong** | Click **Retry** to send your message again. |
## Chat History
Open **AI Marketer → Chat History**. The page, titled **Chats**, lists the conversations in this workspace, most recently updated first. Use **Search chats** to find one.
* **Chats** — your conversations.
* **Agent tasks** — jobs the agent ran on its own, with a status: **Queued**, **Running**, **Done** or **Failed**.
Hover a row and open its menu to **Rename** or **Delete** it. **New chat** starts a new conversation.
Chats created by [scheduled tasks](/agent/schedules), by applying [approvals](/agent/approvals), from Slack and by agent tasks open read-only ("This chat is read-only."). They're a record of what the agent did — to continue, start a new chat.
## Related
What each message and agent run costs.
Ready-made prompts to start from.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.