Skip to main content

AI Agents

AI Agents are intelligent assistants you can configure and tune to handle conversations, answer questions, and automate everyday tasks for your users. Build one from a template, give it a personality, knowledge, and tools, then deploy it as a chat widget on your website or over the phone.

AI Agents

What you can do​

  • Create agents from ready-made templates: Support, Sales, Healthcare, Marketing, Education, Finance, Legal, Recruitment, Technical, Travel, Personal assistant, or a blank Custom starting point.
  • Shape each agent's personality with a name, first message, base prompt, languages, and an attached knowledge base.
  • Turn on built-in System tools so an agent can perform actions, or add your own callable tools to extend it.
  • Choose how a voice agent talks using Voice Playbooks, with control over the speech-to-text, text-to-speech, and realtime models, plus voice, language, and pace.
  • Test an agent live in a side-by-side preview before you publish it.
  • Deploy as an embeddable Chat widget or assign a Phone number for incoming calls.
  • Watch performance on the dashboard, including active sessions, top intents, and a fallback report, and review every past conversation in History.

How it works​

When you create an agent, RedFlare stores it as a draft with the template's personality, your greeting, and a default chat-widget configuration. Nothing is live yet, so you can edit freely.

Here's what happens behind the scenes as you build and ship an agent:

  1. Personality and knowledge. Your prompt and first message define how the agent behaves. If you attach a knowledge base, the agent doesn't just read your documents word for word. Your articles and sources are turned into a searchable, vectorized index, and at conversation time the agent retrieves the most relevant passages to ground its answers, so replies stay accurate and on-topic instead of guessing.
  2. Tools. Each tool you enable is registered as an action the agent can call mid-conversation when it decides one is needed, for example looking something up, transferring a call, or sending data to another system.
  3. Voice pipeline. For voice agents, your chosen Voice Playbook decides how speech is handled. A staged pipeline transcribes the caller's speech to text, sends it to the language model that writes the reply, then converts that reply back to speech, while a realtime playbook runs the whole exchange end-to-end for the lowest latency.
  4. Deployment. When you deploy, RedFlare provisions a live runtime for the agent. The agent's status moves to Queued while that happens, then flips to Live once it's ready. RedFlare keeps checking the status in the background and lets you know the moment deployment finishes.
  5. Conversations and insights. Every chat or call becomes a session with a full transcript and a status, such as ongoing, completed, or closed by the user. From these sessions RedFlare derives the dashboard metrics, the top intents people ask about, and the fallback report of questions the agent struggled with.

Create an agent​

  1. Open AI Agents from the sidebar.
  2. Click New Agent. If you don't have any agents yet, you'll see "No Agents yet" with a button to Create new Agent.
  3. Enter an Agent name.
  4. Under Choose a template, pick the type that fits your use case. Each template shows a short description of how it behaves, from an empathetic Support specialist to an energetic Sales consultant, or choose Custom for a blank slate.
  5. Click Create.

Your new agent opens straight to its detail view, ready to configure.

tip

The template only sets a starting personality. You can always rewrite the prompt and first message afterwards, so don't overthink the choice.

Configure your agent​

Open an agent and go to the Configure tab. It's split into three areas: General, Tools, and Models.

General​

  1. Set the agent's Name, the display name users see during interactions.
  2. Choose a Timezone for time-based context.
  3. Set the primary Language, and add any Additional languages the agent can understand or respond in.
  4. Attach a Knowledge base so the agent can refer to your documents and sources when answering. Use "Select a knowledge base" to pick a collection.
  5. Write the First message, the greeting that welcomes or onboards users.
  6. Fill in the Prompt with base instructions that shape the agent's personality and behavior.

Changes are tracked as you type. When you have edits waiting, save them from the unsaved-changes banner before moving on.

note

Attaching a knowledge base enables retrieval for that agent. Keep your collections tidy and up to date in the Knowledge area, since the agent answers from whatever you've published there.

Tools​

Tools give your agent extra abilities beyond conversation.

  1. In Tools, browse System tools, the built-in actions grouped under Conversation, Telephony, Integration, and Other.
  2. Toggle a tool on to enable it. A confirmation appears naming the tool you just enabled. The "What this agent can do" panel lists everything that's currently turned on.
  3. Select a tool to open its configuration and adjust its settings, then save.
  4. To extend the agent further, click Add tool and set up a callable webhook or client-side integration.

If nothing is enabled yet, you'll see "No tools enabled" with a nudge to "Turn on a tool below to give this agent new abilities."

tip

Webhook tools connect to your backend systems, while client tools interact with the browser in real time. Give each tool a clear description so the agent knows exactly when to use it.

Models​

For voice agents, the Models area lets you pick a Voice Playbook and tune the underlying models.

  1. Under Choose Playbook, select a speech flow:
    • STT LLM TTS for a cost-effective, highly controllable pipeline.
    • Realtime TTS for low-latency speech from text.
    • Realtime for near-instant, end-to-end voice conversations.
  2. Configure the relevant models for your playbook:
    • LLM, the brain that understands and writes replies.
    • Speech to Text, which turns what the caller says into text.
    • Text to Speech, which gives your agent its voice.
    • Realtime, which powers instant, end-to-end voice conversations.
  3. Pick the Language, the Agent Voice, and the Pace that controls how fast the agent speaks.

Test your agent​

  1. From the agent header, click Test Agent.
  2. A live preview opens in a side pane where you can chat with the agent exactly as your users would.
  3. Close the pane when you're done.
note

Testing requires the agent to be deployed and live. If the Test Agent button is unavailable, deploy your agent first.

Deploy your agent​

When you're happy with the setup, make the agent live.

  1. Open the Actions menu in the agent header.
  2. Choose Deploy agent. If the agent is already live and you've made edits, you'll see Redeploy agent instead to apply your changes.
  3. Wait for deployment to finish. The status moves through Queued, then turns to Live, and you'll get an "Agent deployed successfully" confirmation.

An agent's status helps you track where it stands: Draft, Queued, Live, Idle, or Error.

caution

You can't deploy an agent while it's Queued. Wait until the previous deployment completes before trying again.

Set up the Chat widget​

  1. Go to the Deployment tab and open Chat widget.
  2. Customize the Appearance: choose a Theme (light, dark, or system), Size (small, medium, or large), Position on the page, and Colors.
  3. Optionally turn on Terms and conditions so users accept them before chatting.
  4. Set your Preferences, including the Agent mode for how users interact: audio only, text only, or both audio and text.
  5. Copy the Embed code snippet and add it to the page where you want the widget to appear.

Set up Phone​

  1. In the Deployment tab, open Phone.
  2. Assign a phone number your agent can receive incoming calls on using "Select a phone number".
note

Phone numbers come from the numbers you've connected through a telephony provider. If the list is empty, add a number first, then return here to assign it.

Review performance and history​

The Dashboard tab shows how your agent is doing once people start interacting:

  • Headline Metrics like active sessions, total conversations, average conversation duration, and user satisfaction rating.
  • Top intents, the most frequently triggered user goals, to help you optimize key conversation paths.
  • A Fallback report that highlights queries the agent couldn't answer well, so you know where to improve.
  • Recent conversations for a quick look at the latest sessions.

Open the History tab to browse every session by mode, duration, message count, and status. Select a conversation to read its full transcript, and download it if you need a copy. If the agent is brand new, you'll see "No conversation history" until it has its first interaction.

tip

Empty top-intents and fallback panels are normal for a fresh agent. Start a few conversations in the test preview to see these insights begin to populate.