Chat — Talking to Your AI Agents
Access: All users (for agents they have been assigned to)
The Chat module is the primary interface for interacting with your AI agents. It provides a familiar, messaging-style experience with enterprise features like real-time streaming, security labels, and session management.
What is the Chat Module?
Think of it as a chat app (like Slack or WhatsApp) purpose-built for AI agents. Instead of chatting with people, you're chatting with your registered AI agents. Every conversation is:
- Streamed in real-time — responses appear word by word as the agent generates them
- Automatically saved — you can return to any conversation at any time
- Secured — AgentGuard monitors every message for safety (if enabled)
- Tracked — token usage, latency, and costs are recorded per message
Demo Video
How to Start a Conversation
Step 1: Find Your Agent
- Click Agents in the sidebar.
- You'll see all agents you have access to displayed as cards (grid view) or rows (list view).
- Each agent card shows:
- Name and description
- Framework badge (ADK, LangGraph, CrewAI) with color coding
- Deployment type (Cloud Run, Agent Engine, Custom)
- Status indicator — a pulsing green dot means "Live", grey means "Offline"
- Tags for categorization (e.g.,
production,support)
Step 2: Open the Chat
- Click the Chat with Agent button on the agent card.
- You'll be taken to the full-screen chat interface.
Step 3: Send a Message
- Type your message in the text input at the bottom.
- Press Enter or click the send button.
- The agent's response will stream in real-time — you'll see it appear word by word.
Expected result: The agent responds with formatted text. Responses can include:
- Plain text paragraphs
- Markdown formatting — tables, code blocks (with copy buttons), bold, italic, lists
- Citations — if the agent references external sources
Managing Chat Sessions
What is a Session?
A session is a single conversation thread with an agent. Each session maintains its own conversation history, so the agent remembers what you discussed within that session.
Creating a New Session
- Click + New Chat in the chat sidebar to start a fresh conversation.
- Each new session is isolated — the agent starts with no memory of previous sessions.
Switching Between Sessions
- The left panel of the chat interface shows all your sessions for the current agent.
- Each session displays:
- Session name (auto-generated from your first message)
- Last message timestamp (shown as relative time, e.g., "2 hours ago")
- Active session highlighting — the current session is visually highlighted
- Click any session to switch to it and see the full conversation history.
Deleting Sessions
- Hover over a session in the sidebar and click the trash icon to delete it.
- Deleted sessions cannot be recovered.
What You'll See in Each Message
Every agent response in TraptureIQ includes metadata for transparency:
| Label | What It Shows | Why It Matters |
|---|---|---|
| AgentGuard Status | "Safe" or "Sanitized" — indicates whether the firewall processed the message | Know instantly if content was modified for safety. If a response is blocked, you'll see a warning message instead. |
| Token Count | The exact number of tokens consumed by the message (input + output) | Track per-message cost and understand how verbose your agent is |
| Latency | How many seconds the agent took to respond | Monitor response times and identify slow agents |
Features at a Glance
| Feature | Description |
|---|---|
| Real-time Streaming | Responses appear word by word as the model generates them — no waiting for the full response |
| Markdown Rendering | Full GitHub-flavored markdown support — tables, code blocks, headings, lists, bold, italic |
| Code Blocks with Copy | Code snippets have a one-click copy button for easy use |
| Session Persistence | Every conversation is saved. Close the browser and come back — your history is preserved |
| File Attachments | Attach files to your messages when supported by the agent |
| Multi-Session Support | Chat sidebar lets you manage multiple conversation threads per agent |
| Auto-Naming | Sessions are automatically named based on your first message |
| Error Handling | If an agent fails to respond, you'll see a clear error message instead of a blank response |
Tips for Beginners
- Start simple — Send a short greeting to test that the agent is working before asking complex questions.
- One agent at a time — The chat interface is scoped to one agent per conversation. To chat with a different agent, go back to the Agents page and select it.
- Check the status — If an agent's card shows "Offline", it means the underlying service is down. Contact your admin or check the deployment.
- Use sessions wisely — For different topics, start a new session. Mixing topics in one session can confuse the agent's context.