Chat with Your Atmet AI Agent to Build Automations

Use natural language to build, manage, and monitor automations — chat with your Atmet AI agent and it handles the configuration for you.

The Atmet agent chat is your primary interface for everything on the platform. Instead of clicking through menus or filling out configuration forms, you describe what you want in plain language and the agent handles the rest — asking clarifying questions, rendering interactive components inline, and building automations on your behalf.

Starting a chat

To open a new conversation, select New Chat from the main navigation sidebar. Each chat creates an isolated session where your agent remembers the full context of that conversation. You can return to any previous chat at any time from the chat history panel.

1. Open a new chat

   Click New Chat in the left sidebar. A fresh conversation window opens with your agent ready to receive instructions.

2. Describe your goal

   Type what you want to accomplish in plain language. You don’t need to use special syntax — just explain the outcome you’re after, as you would to a colleague.

3. Answer the agent's follow-up questions

   The agent may ask clarifying questions or render inline components (forms, tables, file pickers) to collect the details it needs. Respond directly in the chat.

4. Review and confirm

   Once the agent has enough context, it will summarize what it plans to do. Confirm to proceed, or adjust your instructions before the action runs.

How to describe what you want

You don’t need technical knowledge to get results. Speak to the agent the way you’d brief a new team member — focus on the outcome, not the steps.

Effective examples:

• “Every time a new lead fills out our contact form, add them to HubSpot and send a welcome email from Gmail.”
• “Pull last week’s sales numbers from Google Sheets and post a summary to our #revenue Slack channel every Monday morning.”
• “When a customer replies to a support ticket in Jira, draft a response and ask me to approve it before sending.”

The more context you provide upfront — including which apps are involved and what the expected outcome looks like — the fewer follow-up questions the agent will need to ask.

Generative UI: interactive components in chat

Rather than responding with plain text alone, the Atmet agent renders interactive components directly inside the conversation to collect structured information or display results.

Forms

The agent presents input fields when it needs structured data from you, such as email addresses, dates, or configuration values. Fields can be text, email, textarea, select, date, time, or number.

Tables

Query results, contact lists, and data previews appear as scannable tables so you can review information before the agent acts on it. Tables can display up to 100 rows.

Charts

Bar and line charts surface trends and summaries — for example, when you ask the agent to analyze your pipeline data or visualize weekly metrics.

Code blocks

Scripts, API responses, and generated configurations are displayed in syntax-highlighted code blocks with an optional filename, so you can copy or reference them easily.

File lists

When the agent retrieves documents from Google Drive or other sources, it presents them as a browsable file list showing the file name, type, owner, and last-updated date.

Generative UI components are fully interactive. You can fill out a form, select rows in a table, or choose a file — all without leaving the chat window.

Mentioning integrations with @syntax

To tell the agent which app or integration to use, prefix the integration name with @. This pins the agent’s attention to a specific connected service and speeds up configuration.

Examples:

• @gmail — send or read emails through your connected Gmail account
• @slack — post messages or read channels in your Slack workspace
• @hubspot — create or update contacts and deals in HubSpot
• @telegram — send messages through a connected Telegram bot
• @googlesheets — read from or write to a specific spreadsheet

You can mention multiple integrations in a single message: “Watch @gmail for invoices and log them to @googlesheets.”

Integrations must be connected in your workspace settings before you can mention them. Visit the Integrations section in the sidebar to link your accounts.

Attaching files for context

You can attach files directly to a chat message to give the agent additional context — for example, a sample CSV to define a data format, a PDF contract to extract terms from, or a template document to replicate.

To attach a file, click the paperclip icon in the chat input bar and select a file from your computer, or drag and drop it into the message field. The agent will acknowledge the attachment and use it as context for its response.

Useful attachment scenarios:

• Upload a spreadsheet so the agent knows your exact column structure before building a data workflow
• Share a brand voice document so the agent writes copy in the right tone
• Provide a sample email thread so the agent understands the communication style you expect

Example conversation

Below is an example of how a typical agent conversation flows from a rough idea to a confirmed automation.

You:       I want to follow up automatically when a deal in HubSpot hasn't
           moved in 14 days. Send a nudge email from @gmail.

Agent:     Got it. To set this up I need a few details:
           [Form rendered inline]
           - Which HubSpot pipeline should I watch?
           - What deal stages should trigger the follow-up?
           - Should I send the email automatically or ask you to approve
             each one first?

You:       [Fills in form] Use the "Enterprise Sales" pipeline, any deal
           in "Proposal Sent" or "Negotiation". Ask for approval first.

Agent:     Here's what I'll build:
           • Trigger: Every day at 9 AM, check for deals in "Proposal Sent"
             or "Negotiation" with no activity in 14+ days.
           • Action: Draft a follow-up email and pause for your approval
             before sending via @gmail.

           Ready to activate this workflow?

You:       Yes, activate it.

Agent:     ✓ Workflow "HubSpot Deal Nudge" is now active.

Sharing a chat with teammates

To collaborate on a conversation, open the chat’s Share menu (the icon in the top-right corner of the chat window). You can share a read-only link with anyone in your workspace, or grant edit access so a teammate can continue the conversation and build on your progress.

Shared chats expose the full message history of that conversation, including any data the agent retrieved. Review the contents before sharing with colleagues who shouldn’t see sensitive information.

Managing chat history

All your past conversations are listed in the Chat History panel in the left sidebar, sorted by most recent. Click any chat to reopen it and pick up where you left off — the agent retains full context for the session.

Searching and organizing past chats

Use the search bar at the top of the chat history panel to find a conversation by keyword. You can also rename any chat for easier reference by clicking the three-dot menu next to the chat name and selecting Rename.

Deleting a chat

To remove a conversation, open the three-dot menu next to the chat name and select Delete. Deleting a chat does not delete any workflows or automations that were created during that session — those live independently in your Workflows section.