Testing Your Agent

Test your agents thoroughly before deploying to production.

Testing Methods

Test Panel (In-Editor)

The fastest way to test while building:

1. Open your agent in the editor
2. Click the Test tab in the right panel
3. Start a conversation with your agent
4. See responses in real-time
5. View conversation logs and debug info

Features:

• Real-time interaction
• View agent's reasoning
• Debug function calls
• No phone number required
• Best practice: keep warm-up enabled to reduce first-turn latency (the first test run may take slightly longer while models/connectors prime).

Web Demo

Test with voice in your browser:

1. Navigate to your agent in the Agents list
2. Click Test in Browser or the play button
3. Allow microphone access
4. Speak to your agent
5. Listen to responses

This simulates a real voice call without using a phone number.

Test Calls

Make actual phone calls to test your agent:

1. Purchase or connect a test phone number
2. Assign your agent to the number (inbound)
3. Call the number from your phone
4. Experience the conversation as users will

Recording Playback

Review past test conversations:

1. Open the agent editor
2. Go to the Test tab
3. Find previous test sessions
4. Click to play back recordings
5. Review transcripts

What to Test

Basic Functionality

Conversation Flow

For Conversation Flow agents, test each path:

• Happy Path - The expected conversation flow
• Alternate Paths - Different user responses
• Edge Cases - Unexpected inputs
• Transitions - Moving between flow nodes

Functions

Test all configured functions:

1. Trigger each function during conversation
2. Verify correct parameters are passed
3. Check response handling
4. Test error scenarios (API failures, etc.)

Voice Quality (Voice Agents)

Test audio characteristics:

• Clarity - Is the voice clear and understandable?
• Pacing - Is the speech speed appropriate?
• Interruptions - How does it handle being cut off?
• Background Noise - Test with ambient noise

Testing Checklist

Before First Deploy

• Greeting works correctly
• All conversation paths tested
• Knowledge base queries return relevant results
• Functions execute successfully
• Error handling is graceful
• Voice sounds natural (voice agents)

After Changes

• Changed functionality works as expected
• Existing functionality still works (regression)
• New edge cases handled
• Performance is acceptable

Iterating Based on Tests

Document Issues

Track problems you discover:

• What scenario caused the issue?
• What was the expected behavior?
• What actually happened?
• Screenshots/recordings if helpful

Prioritize Fixes

Focus on issues by impact:

Fix and Re-test

1. Update agent configuration
2. Test the specific fix
3. Run regression tests on related functionality
4. Deploy when confident

Common Issues and Solutions

Agent Doesn't Understand Questions

Causes:

• Prompt instructions too vague
• Missing context in knowledge base
• Unexpected phrasing

Solutions:

• Add example phrases to prompts
• Expand knowledge base content
• Add synonyms and variations

Responses Are Too Long/Short

Causes:

• Prompt instructions don't specify length
• Agent trying to be too helpful/brief

Solutions:

• Explicitly set response length in prompts
• Add examples of ideal response length
• Adjust verbosity settings

Knowledge Not Being Used

Causes:

• Knowledge base not connected to agent
• Content not matching queries
• Retrieval threshold too strict

Solutions:

• Verify knowledge base is linked in agent settings
• Improve knowledge base content structure
• Adjust retrieval sensitivity

Functions Not Triggering

Causes:

• Function trigger conditions too narrow
• Missing parameters in conversation
• API configuration errors

Solutions:

• Broaden trigger conditions
• Adjust prompts to gather required info
• Test API endpoint independently

Conversations Feel Robotic

Causes:

• Overly formal prompts
• No personality defined
• Lack of conversational phrases

Solutions:

• Add personality traits to prompts
• Include small talk capabilities
• Add filler phrases and acknowledgments

Best Practices

Testing Strategy

1. Test Early - Don't wait until "done"
2. Test Often - After every significant change
3. Test Systematically - Use a checklist
4. Test Realistically - Use actual user scenarios

Involving Others

1. Internal Testing - Team members test first
2. Beta Testing - Limited external testing
3. Feedback Collection - Gather structured feedback
4. Iteration - Incorporate feedback before full launch

Environment Management

1. Development Agent - Test changes freely
2. Staging Agent - Final testing before production
3. Production Agent - Live with real users

Use separate agents or duplicate/export to manage versions.

Next Steps

• Deploy to production
• Set up batch calls
• Monitor performance