Agent Definition
This guide provides a comprehensive walkthrough for creating and configuring AI agents in FutureAGI’s simulation platform. Agent definitions form the foundation of your voice conversational testing.What is an Agent Definition?
An agent definition is a configuration that specifies how your AI agent behaves during voice conversations. It includes:- Basic information (name, contact details)
- Provider settings (AI model configuration)
- Agent description (prompts and behavior guidelines)
- Communication preferences
Creating an Agent Definition
Step 1: Access Agent Definitions
Navigate to Simulations → Agent Definitions in your FutureAGI dashboard. Click the “Add Agent Definition” button to begin creating a new agent.
Step 2: Configure Basic Information

Basic Information
- Agent Name:
Insurance Sales Agent
- Agent Type: Choose between ‘voice’ or ‘chat’
Voice Configuration
- Provider: Select provider (e.g., ‘Vapi’ or ‘Retell’)
- Assistant ID: Enter your provider-specific assistant ID (optional)

Behavior Configuration
- Prompts/Chains: Define your agent’s behavior, personality, and conversation flow. This field accepts your agent’s prompts and operational guidelines.
Language Selection
Choose the primary language for your agent:- English (en)
- Spanish (es)
- French (fr)
- German (de)
- [Additional languages based on provider]
Knowledge Base
Provide the LLM with domain-specific information to help it answer questions more accurately. Upload relevant documents, FAQs, or product information to enhance your agent’s expertise.
Contact Information
- Contact Number:
+1-800-INSURE-ME
(or your test number) - Pin Code: Select your country code (e.g., +1 for US)
- Connection Type: Toggle to
Inbound
(agent receives calls)
Version Control
Commit Message: Add a descriptive message for each configuration change to track updates and maintain version history.- Example: “Updated greeting script to include new product line”
- Example: “Modified objection handling for premium concerns”
Connection Settings
Connection Type
Inbound (Toggle ON):- Agent receives incoming calls
- Suitable for customer service, support hotlines
- Agent waits for customer to initiate
- Agent makes outgoing calls
- Suitable for sales calls, appointment reminders
- Agent initiates the conversation
Understanding Agent Description
An Agent Description is a comprehensive overview of your voice conversational agent’s purpose, behavior, and operational guidelines. It encapsulates all the instructions, prompts, and rules you provide to your AI agent, acting as the foundational reference for how the agent should interact with users and handle various scenarios. Why it matters: By clearly articulating your agent’s description, you enable FutureAGI to:- Generate more accurate test scenarios
- Identify issues in production
- Ensure consistent agent behavior
- Validate compliance and quality standards
Agent Description Templates
Single Prompt Agent
For agents with a single, unified prompt, simply copy your agent’s main prompt into the description field. Example - Insurance Sales Agent:Multi-State Agent
For agents with multiple conversation states or nodes, use this structured format:Retell Workflows
If you’re using Retell AI for your agent, export and paste the configuration:- Go to Agents in Retell dashboard
- Select your agent
- Click the Export button (downloads a .json file)
- Copy the entire JSON content
- Paste it in the agent description field
Vapi Workflows
For Vapi-based agents, export the workflow configuration:- Go to Workflows in Vapi dashboard
- Select your agent workflow
- Click the Code button (top right)
- Copy the displayed content
- Paste it in the agent description field
Best Practices for Agent Descriptions
1. Be Comprehensive
Include all relevant information:- Personality and tone
- Product knowledge
- Compliance requirements
- Objection handling strategies
- Escalation procedures
2. Use Clear Structure
Organize your description logically:- Start with role and personality
- List key responsibilities
- Include do’s and don’ts
- Add specific examples
3. Include Edge Cases
Specify how to handle:- Angry or frustrated customers
- Technical questions beyond scope
- Requests for human agents
- System errors or unknowns
4. Add Compliance Rules
For regulated industries:- Required disclosures
- Prohibited statements
- Documentation requirements
- Privacy guidelines
5. Test Scenarios
Consider including:- Common customer questions
- Typical objections
- Success criteria
- Failure conditions
Advanced Configuration Options
Custom Variables
You can include variables that the simulation engine will use:Behavioral Modifiers
Add specific behavioral instructions:Integration Hints
If your agent integrates with systems:Validation and Testing
After creating your agent definition:- Review All Fields: Ensure accuracy
- Test Description: Check for completeness
- Save as Draft: If still refining
- Create Agent: When ready for testing
Common Mistakes to Avoid
1. Vague Descriptions
❌ “Be helpful and friendly” ✅ “Greet with ‘Good [morning/afternoon], thank you for calling SecureLife. I’m Sarah, your insurance specialist. How may I assist you today?‘“2. Missing Compliance
❌ No mention of regulations ✅ “Always state: ‘This call may be recorded for quality and training purposes’ at the beginning”3. Incomplete Product Info
❌ “Sell insurance products” ✅ “Offer: Term Life (10/20/30 year), Whole Life, Universal Life. Minimum coverage: $50,000”4. No Error Handling
❌ No guidance for problems ✅ “If unable to answer: ‘That’s a great question. Let me connect you with a specialist who can provide the most accurate information.’” [10:58] Raj Shekhar SinhaAgent Definition Details
Introducing Agent Definition Versioning
Agent definition versioning allows you to track changes made to your AI agents over time. Each version captures the agent’s configuration, behavior prompts, knowledge base connections, and other key settings. With versioning, you can safely experiment with updates, roll back to previous versions, and maintain an audit trail of your agent development.Understand the UI
The Agent Details UI is divided into key sections:- Agent Select Dropdown – Switch between different agents quickly.
- Version Management Section – Located on the left, shows all versions with the latest at the top. Each version displays:
- Version number
- Timestamp
- Commit message
- Create New Version Button – Opens a side drawer to create a new version of the agent.

How Versioning Agents Helps You
Versioning provides several benefits:- Experiment Safely – Test new prompts, workflows, or provider settings without affecting the live agent.
- Rollback Capability – Restore any previous stable configuration if needed.
- Audit & Compliance – Maintain a history of agent modifications for regulatory or internal compliance.
How to Create New Agent Versions
When creating a new version:
- Click Create New Version in the version management section.
- In the side drawer, complete:
- Commit Message – Describe the changes
- Basic Information – Agent name, description, etc.
- Configuration Fields – Behavior, voice, and knowledge base
- ✅ Click Save to create the version.
💡 Tip
Always provide clear commit messages to make version history meaningful.
Switching Between Versions

- In the Version Management section, click any existing version.
- The UI will load the selected version for viewing, configuration, and further edits.
- This allows users to quickly switch between different configurations of the same agent.
⚠️ Note Switching versions does not delete previous versions; all historical versions remain accessible.
Exploring Different Tabs
Agent Configuration Tab ⚙️

- Basic Information
- Voice Configuration
- Behavior Configuration
- Contact Information
- Associated Knowledge Base
Performance Analytics Tab 📊

- Call success rates
- Average response times
- Evaluation scores across multiple metrics
- Error rates and anomalies
- ✅ Identify strengths and weaknesses in agent behavior
- 📈 Monitor improvements over time
- 🚨 Quickly spot issues in production or testing
Call Logs Tab 📞

- Call Information – Duration, participants, and call status (Completed, Failed, Dropped)
- Evaluation Scores – Scores for each call on defined metrics
- Call Details Drawer – Click any call to open:

- Full conversation transcript
- Turn-by-turn analysis
- Evaluation results per metric
- Audio playback (if enabled)
- Key moments flagged by evaluations
Takeaways & Best Practices
Agent versioning empowers you to safely iterate on your AI agents while maintaining a full history of changes. To make the most of versioning:- Always provide clear commit messages to describe changes.
- Test new versions in a staging environment before deploying live.
- Use performance analytics and call logs to monitor improvements and identify potential issues.
- Regularly review historical versions to learn from past configurations and optimize agent behavior.
Next Steps
Once your agent definition is created:- Test with Simple Scenarios: Start with basic test cases
- Refine Description: Based on initial results
- Create Comprehensive Scenarios: Build out test suites
- Run Full Simulations: Execute complete test runs
- Iterate and Improve: Use results to enhance the agent