Using the Agent

Open the Agent

Go to Agent from the main navigation. If your role does not show Agent, ask an administrator to confirm the core/agents:chat permission.

The Agent page includes:

• A conversation history panel.
• A central chat panel.
• A right-side artifact workspace.
• A health banner when system status needs attention.
• Upload, send, stop, share, duplicate, history, and artifact workspace controls.

[picture 1. Agent page with labeled history, chat, input, health banner, and artifact workspace]

Start a new conversation

Click New Conversation or type into the empty Agent screen. The first message becomes the default conversation title unless you rename it later.

Good first prompts:

• "What data do we have available?"
• "What dashboards do I have access to?"
• "Show me all metric definitions and what they measure."
• "Build me a dashboard to track our key performance metrics."

Ask a focused question

The Agent works best when you include:

• Metric or business concept.
• Time period.
• Breakdown.
• Output format.
• Any filters.

Example:

Create a table artifact showing net revenue, order count, and AOV by channel for the last 90 days. Sort by net revenue descending.

Use follow-ups

Follow-ups use the current conversation context. You can ask:

• "Break that down by week."
• "Exclude refunds."
• "Use the same metric but compare to the prior period."
• "Turn this into a chart card."
• "Add the chart to a dashboard."

Start a new conversation when you want a clean context.

Work with selected artifacts

When you select a table, chart card, or dashboard in the artifact workspace, the Agent receives that active artifact context with your next message.

Examples:

• Select a table artifact, then ask: "Add a row explanation for the top five rows."
• Select a chart card, then ask: "Change this to last 90 days."
• Select a draft dashboard, then ask: "Add a tab for customer retention."

Attach existing artifacts

You can attach existing dashboards or artifacts to the current conversation. If no conversation is open, Stackless can start a new conversation with the selected artifacts.

Use this when:

• You want to continue a prior analysis.
• You want the Agent to modify a prior dashboard draft.
• You want to compare a saved table artifact with a new question.
• You want to build a dashboard from existing chart cards.

[picture 2. Existing artifact picker with selected dashboard and table artifact]

Upload files

Use the paperclip icon to upload CSV or XLSX files. Uploaded files are staged for the next message and can persist with the conversation.

Tell the Agent what the file represents:

• "This workbook is seed data for a new analysis."
• "Compare this CSV to existing warehouse orders."
• "This sheet should be matched to the existing customer source."

See Uploads for details.

Stop a response

Use the stop button when a response is still streaming or thinking. If you switch conversations or start a new conversation while the Agent is running, Stackless asks you to confirm because the unfinished response may be lost.

Some backend operations can continue after the chat response is interrupted. Check Monitor or the artifact workspace if you think a draft, artifact, or operation was already created.

Share and duplicate conversations

Conversation owners can share a conversation. Shared conversations are read-only for recipients. A recipient can duplicate a shared conversation to create an editable copy.

Do not use sharing as an access workaround. Dashboard, schema, artifact, and role permissions still apply.

When to ask for a durable output

Ask for a durable output when the result should be reused:

• "Save this as a table artifact."
• "Create a chart card."
• "Write a brief with key metrics and supporting evidence."
• "Create a finding card with confidence and next action."
• "Build a dashboard from these results."

If you only need a quick answer, a chat response may be enough. If the result will be inspected, exported, shared, or repeated, make it an artifact or dashboard.