Chat with an agent

The App is where you actually use the agents an admin built for your workspace — you hold conversations with them, fill forms through chat, and run extractions. It’s the everyday, end-user side of the platform (Studio is the separate place where agents are configured).

You open the App from your workspace card on the home page — its button is labelled App (grid icon). Everyone with access to a workspace can open it.

How it works

• The App’s left sidebar lists every agent in the workspace; the main area shows the same agents as cards. You pick one to talk to.
• Each agent keeps its own sessions (past chats). You start a New Chat or reopen a previous session — they’re saved and listed under the agent.
• Agents come in three types, and the App adapts to each:
  • Conversation — a normal chat.
  • Form — a chat that also fills a Form Result panel as you answer.
  • Extraction — instead of chatting, you run an extraction on a document.

The full flow at a glance

The walkthrough below replays every step in the real interface. Use Prev / Next to move at your own pace; each step highlights the button to click and the area to watch.

Step by step

1. Open the App

On the home page, find your workspace card and click App (the grid icon). The App opens with the workspace’s agents in the left sidebar and as cards in the main area.

2. Pick an agent

Choose the agent you want to talk to — click its card in the main area, or its name in the sidebar. Each agent shows its type (Conversation, Form, or Extraction).

3. Start or reopen a session

On the agent’s page, click New Chat (New Form Session for a form agent, or New Extraction for an extraction agent) to begin. Your past sessions are listed alongside — click one to reopen it. In the sidebar, an agent’s sessions appear nested under it, newest first.

4. Send a message

Type your question in the box at the bottom (Ask your question…) and press Enter (or the send button) to send it. The agent’s reply streams into the thread just below your message. Use Shift + Enter for a line break.

5. Use the composer tools

Next to the input you’ll find:

• Attach (paperclip) — add a file (e.g. a PDF or image) to your message.
• Voice input (microphone) — dictate instead of typing (shown only if your browser supports it).
• Jump to a message (list icon) — in longer threads, quickly jump back to any of your earlier questions.

6. Work with an answer

While the agent works, a live status shows what it’s doing (e.g. Working through 2 steps); once done it collapses to a Worked through … steps summary you can expand to see each step. Under each finished answer you can:

• Feedback — a thumbs button opens Report a problem with this message to send a comment to the team.
• Copy — copy the answer to your clipboard.
• Sources — when the agent used your workspace’s documents, a … sources button opens a panel listing them (with a download link for public documents). This appears only when the workspace has the sources feature enabled.

7. Form agents: watch the Form Result

A form agent uses the same chat, plus a Form Result panel on the right. As you answer the agent’s questions, the panel fills in each field; fields not yet collected show a —. On a narrow screen the panel becomes a collapsible strip above the chat.

Managing sessions

• Reopen any past session from the agent’s list or the sidebar — everything is saved.
• Delete a session with the trash icon (on the session card, in the sidebar’s session menu, or in the chat header). There is no rename.

Extraction agents

An extraction agent doesn’t chat. On its page you click New Extraction, pick or upload a document (PDF/image, or a CSV for batch runs), and read the extracted result on a dedicated run page. Its runs are listed as history rather than chat sessions.

Tips

• Give the agent context in your first message — the clearer the question, the better the answer.
• Use Sources to check where an answer came from before relying on it.
• Leave feedback on a weak answer; it helps the team improve the agent.
• Start a New Chat for a new topic so each conversation stays focused.

Troubleshooting

• I don’t see any agents — no agents have been created in this workspace yet, or you don’t have access. Ask a workspace Admin (they build agents in Studio).
• The voice button is missing — dictation only appears when your browser supports speech recognition; try a recent Chrome or Edge.
• There’s no Sources button — the agent either didn’t use any documents for that answer, or the workspace doesn’t have the sources feature enabled.
• I can’t rename a session — renaming isn’t available; sessions are labelled by their time. You can delete a session you no longer need.
• I opened Studio by mistake — Studio is the admin/configuration side. If you only want to use the agents, go back to the App.