Agent Builder - Basic Configuration¶
Essential configuration steps to get your AI agent up and running quickly.
Overview¶
This guide covers the minimum required configurations to create a functional AI agent. You can complete these steps in 10-15 minutes to have a working agent.
For advanced features like voice processing, custom tools, and personality traits, see Agent Builder - Advanced Configuration.
Prerequisites¶
Before you begin, ensure you have:
- Access to https://PLATFORM-URL-PLACEHOLDER platform
- Appropriate permissions to create new agents
- Basic understanding of your agent's purpose
Quick Start: 4 Essential Steps¶
To create a functional agent, you need to configure:
- Agent Info - Name, type, and where it will be used
- Instructions - What your agent does and how it behaves
- Models - Which AI model powers your agent
- Context Processing - How your agent handles large conversations
That's it! These four configurations are sufficient to launch a working agent.
1. Agent Info Configuration¶
When to use
Configure these settings when creating any new agent. This defines your agent's identity, deployment channels, and access controls.
Overview¶
The Agent Info section defines the basic properties, deployment channels, and security settings for your agent.
Fields¶
| Field | Required | Description |
|---|---|---|
| Name | Yes | Internal agent identifier used in the platform, logs, and management (e.g., "Customer Support Agent") |
| AI Name | Yes | Short display name shown to end-users within the chat interface. Should be concise and reflect the agent's purpose (e.g., "Alex", "AI") |
| Description | Yes | Brief description of the agent's purpose and functionality |
| Notes | No | Internal notes about the agent (e.g., "Use the agent as a sidebar to an internal or external website") — not visible to end users |
| Agent Type | Yes | Select the agent type: Chat (conversational agent) or Search (search-focused agent) |
| Template | No | Set visibility: Workspace Public (visible to all workspace members) or Unlisted |
| Agent Icon | No | Select an icon from the dropdown. Displayed before the agent's name in the chat interface |
Modalities¶
Select the channels through which users will interact with your agent. You can enable multiple modalities simultaneously.
| Modality | Description | Best For |
|---|---|---|
| Web App | Access the agent directly by entering its address into the browser. Standalone chat interface | Standalone agents that don't need embedding |
| Sidebar | Integrate the agent as a sidebar within an internal or external website. The agent can perform tasks based on the webpage content | Documentation sites, internal portals, content-aware assistance |
| Floating Widget | Classic website chat widget that appears at the bottom right corner of webpages | On-the-spot assistance for website visitors |
Embedding and Testing¶
After saving your agent, the Bookmarklets and Embed on Website options become available (accessible from the three-dot menu on the agent card).
Refreshed Chat UI
The chat experience on Web App, Sidebar, Conversations, and Human Agent screens has been revamped for a sleeker, more conversational layout:
- Agent and user interactions appear aligned to the middle of the screen
- The language-model picker and attach-file option are now inside the message input box
- Stop generation and Regenerate response are now part of the message input box (saving space)
- Prompt shortcuts appear in a sleeker arrangement
Refreshed Web App chat — language-model picker and attach-file inside the message input, prompt shortcuts in a sleek arrangement 
Refreshed Sidebar chat on a host site — same revamped input affordances rendered inside the sidebar panel
The Embed on Website modal provides four sections:
| Section | Description |
|---|---|
| URL | Direct link to the agent's Web App |
| Sidebar | Code snippets to embed the agent as a sidebar on your website |
| Floating Widget | Code snippets to add a chat bubble to the bottom right of your website |
| Bookmarklet | Drag-and-drop bookmarklet for testing the agent on any domain (allowed by CORS) without embedding |
For Sidebar and Floating Widget, two script options are provided:
- Minimal Script (Recommended) — Simple way to add the widget with minimal configuration and default settings
- Extensible Script (Advanced) — Allows advanced customization, overriding default behaviors, passing dynamic user data via
getUserDatafunction
Bookmarklet testing:
- Click and hold the bookmarklet button (e.g., "Agent Name (Floating)" or "Agent Name (Sidebar)")
- Drag it to your browser's bookmarks bar
- Navigate to any domain allowed by CORS
- Click the bookmarklet to load the agent (may take up to 30 seconds)
- Chat controls appear on the page — click the chat icon to open the agent
Security Settings¶
Security settings are split between Web App/Sidebar and Widget, as they support different authentication models.
Web App & Sidebar Security:
| Option | Description |
|---|---|
| Sign-In Required | Users must sign in before they can interact with the agent. Provides user attribution and conversation tracking |
| Sign-In Optional | Users can interact without signing in. Optional login for enhanced features |
Widget Security Settings:
| Option | Description |
|---|---|
| No Authentication | No internal authentication required. Current default mechanism |
| Required Login | Users must complete Internal Authentication to interact with the agent |
| Optional Login | Users can choose to interact in authenticated or unauthenticated mode |
When Required Login or Optional Login is selected, additional fields appear:
| Field | Description |
|---|---|
| Guest User Label | Label displayed on the chat widget for unauthenticated users (e.g., "Login") |
| Logged-in User Label | Label displayed on the chat widget for authenticated users (e.g., "Logout") |
Widget Authentication
Widget authentication uses Internal Authentication (email-based). A working SMTP configuration on the workspace is required for sending authentication codes. When enabled, system tags internal_authenticated and internal_unauthenticated become available for conditional behavior. See Look & Feel for widget customization options.
2. Instructions¶
When to use
Instructions are required for every agent. They define your agent's greeting, role, behavior, and response style. For advanced instruction features (Conditional Instructions, Special Instructions, Current Message Prompt), see Builder — Advanced Configuration.
Overview¶
The Instructions section is where you set up the system instruction and the initial message for your agent. The system instruction acts as a guiding message that steers the agent's responses, essentially serving as a set of directions for how the agent should formulate replies.
Initial Message¶
Enter the text that you want your agent to display as the first message in every conversation. This is usually a greeting or an introductory message.
Examples:
- "Hi! What can I help you with?"
- "Welcome to ChatAgent! Please type your query below."
Agent Instructions¶
Define the instruction that guides the agent's responses. This sets the tone and provides direction for the agent's replies.
- When the agent has no context configured, the Agent Instruction alone will shape responses (creating a general-purpose agent)
- When the agent has context configured (via the Context tool), the agent will work on the context provided
You can use {special_instructions} and {user_instructions} as placeholders in your instructions, which will be substituted with special instructions and user instructions respectively.
What to include:
Role Definition:
You are a customer support agent for fifthelement.ai, specializing in
helping users understand our AI platform features.
Responsibilities:
Your responsibilities include:
- Answering questions about platform features
- Helping users troubleshoot configuration issues
- Providing step-by-step guidance for agent setup
- Escalating complex technical issues when necessary
Behavior Guidelines:
Communication Style:
- Professional yet friendly tone
- Concise and action-oriented responses
- Use bullet points for clarity
- Provide examples when explaining concepts
Constraints:
Limitations:
- Do not provide information about pricing (direct to sales team)
- Do not make promises about future features
- Admit when you don't know something rather than guessing
Advanced Instructions
For conditional instruction blocks, special instructions (Co-Admin role), current message prompt manipulation, and end-of-conversation messages, see Builder — Advanced Configuration.
3. Models¶
When to use
Every agent needs at least one language model configured. Choose based on your use case complexity and performance requirements.
Overview¶
Select the language models that power your agent's intelligence and reasoning capabilities.
Adding Language Models¶
Steps:
- Click "Add Language Model" button
- Click the "Select Language Model" dropdown
- Choose from available models:
- Azure GPT 4.1 Mini
- Gemini 2.5 Flash
- Claude 3.5 Sonnet
- GPT-4
- Click "Save" to apply
Model Selection Guide¶
When to use
- Simple FAQ Agent: Use GPT 4.1 Mini (fast, cost-effective)
- Complex Support Agent: Use Claude 3.5 Sonnet (strong reasoning)
- Technical Documentation Agent: Use GPT-4 (high accuracy)
- Conversational Agent: Use Gemini 2.5 Flash (balanced speed/quality)
| Factor | Description |
|---|---|
| Response Quality | More advanced models provide better reasoning and accuracy |
| Speed | Some models respond faster (e.g., Mini, Flash models) |
| Cost | Different models have different pricing per token |
| Use Case | Match model capabilities to your agent's complexity |
4. Context Processing Algorithm¶
When to use
This setting is required before saving your agent. It controls how your agent handles conversations that exceed the model's context window.
Overview¶
Required Setting: You must select a Large Context Processing Algorithm before saving the agent.
Select the Large Context Processing Algorithm
Available Algorithms¶
Truncate: - Cuts off context when it exceeds model limits - Fast and simple - When to use: Short conversations, simple queries, FAQ agents
Summarize: - Summarizes older context to fit within limits - Maintains conversation continuity - When to use: Long conversations with historical context, customer support
LLM Prompt Chaining: - Breaks down large contexts across multiple prompts - Handles very large contexts - When to use: Complex, multi-step tasks requiring extensive context
Embeddings: - Uses vector embeddings for context management - Efficient retrieval of relevant context - When to use: Knowledge-intensive agents with RAG capabilities
5. Save Your Agent¶
Click Save to finalize your agent configuration
Action: Click "Save" button at the bottom of the page
What this does: - Saves all configurations - Makes the agent available for use - Deploys to selected modalities (Web App, Sidebar, Widget)
Testing Your Agent¶
After saving, test your agent:
- Navigate to the Web App URL (found in agent settings)
- Send a few test queries
- Verify responses match your instructions
- Check that the agent behaves as expected
Next Steps¶
Your agent is now functional! Consider these enhancements:
- Add Knowledge Base: Connect documents or knowledge repositories (Context Configuration)
- Enable Tools: Add capabilities like RAG retrieval or API integrations (Tools Configuration)
- Customize Appearance: Brand your agent's look and feel (Look & Feel)
- Add Shortcuts: Create quick actions for common tasks (Prompt Shortcuts)
- Configure Voice: Enable voice interactions (Audio and Speech)
For all advanced configurations, see Agent Builder - Advanced Configuration.
Related Topics¶
- Agent Builder - Advanced Configuration - Advanced features and power user settings
- Agent Reports - Monitor agent performance and analytics
- Agent's Look & Feel - Customize branding and appearance
- Cloning an Agent - Duplicate and modify existing agents
- Version Control - Manage agent versions and rollback