Skip to main content
Testing is crucial for building reliable workflows. Agility provides multiple testing approaches to help you validate your automations before deployment.

​
Testing Approaches

Agility offers two levels of testing:
  1. Individual Agent Testing - Test a single agent in isolation
  2. Workflow Testing - Test the entire workflow from start to finish
Always test individual agents before testing the complete workflow. This helps identify configuration issues quickly.

​
Testing Individual Agents

1

Open Agent Configuration

Click on the agent card you want to test. The configuration sidebar opens on the right side with a colored header matching the agent type.
2

Configure the Agent

Complete the agent configuration:
  • Add required API keys
  • Set up prompts or parameters
  • Configure any input field references
  • Select appropriate models or settings
Each agent type has different configuration requirements. Check the configuration form for required fields.
3

Save Configuration

Click the “Save Configuration” button (green button). You’ll see a confirmation message: “Configuration saved successfully!”This saves your settings so they’re available when running the workflow.
4

Run Agent Test

Click the “Test Agent” button (blue button) to test the agent individually. The button will show:
  • “Loading…” while the test runs
  • A spinning icon during execution
  • Results or errors when complete
5

Review Results

Test results appear in the configuration sidebar:
  • Success: Generated output is displayed in a gray box
  • Error: Error messages appear in a red box with details
For Text Generator agents, you’ll see the generated text. For other agents, you’ll see relevant output.

​
Test Mode Features

When testing individual agents:
  • Tests run in isolation without executing connected agents
  • Input data can be simulated or left empty
  • No data is saved or sent to external services (depends on agent type)
  • API calls are made with your configured keys
  • Results are logged for debugging
Some agents (like Gmail Sender or Discord Messenger) will actually send messages when tested. Use test accounts or channels when testing these agents.

​
Testing Complete Workflows

1

Configure All Agents

Before testing the workflow, ensure every agent is properly configured:
  • API keys are set
  • Prompts include proper field references
  • All required parameters are filled
  • Each agent has been individually tested
2

Save the Workflow

Click “Save Workflow” in the top-right corner. The button shows:
  • “Saving…” with a spinner during save
  • Returns to “Save Workflow” when complete
A toast notification confirms: “Workflow saved successfully!”
3

Select Starting Agent

Click on the agent where you want to start the workflow execution. This is typically:
  • An input agent (Gmail Reader, GitHub Reader)
  • The first agent in your workflow sequence
4

Run Workflow Test

In the agent’s configuration sidebar, click the “Test Workflow” button (purple button). The button displays:
  • “Running Workflow…” with animated spinner
  • Progress indication during execution
5

Monitor Execution

Watch for toast notifications during execution:
  • “Running workflow…” (info) when starting
  • “Workflow completed successfully! (X steps executed)” (success) when done
  • Error messages if something fails
6

Review Results

Workflow results are stored in localStorage for debugging:
  • Open browser DevTools
  • Go to Application/Storage → Local Storage
  • Find lastWorkflowResults key
  • View JSON data with results from each step

​
Workflow Execution Flow

When you test a workflow:
  1. Starting point: Execution begins at the selected agent
  2. Sequential processing: Agents execute in order based on connections
  3. Data passing: Output from each agent flows to connected agents
  4. Error handling: Execution stops if any agent fails
  5. Result collection: All outputs are collected and returned

​
Understanding Test Results

​
Success Indicators

Individual Agent Test: 
Generated Text
─────────────────────────────────
[Your generated output appears here]
Workflow Test: 
Workflow completed successfully! (3 steps executed)

​
Error Messages

Configuration Errors:
  • “API key is required”
  • “You must be logged in to save configuration”
  • “Failed to save configuration”
Execution Errors:
  • “Failed to generate text”
  • “You must be logged in to test the agent”
  • “Workflow failed: [specific error message]”
Network Errors:
  • “Error: 401” - Authentication failed
  • “Error: 429” - Rate limit exceeded
  • “Error: 500” - Server error
Error messages include specific details to help you diagnose issues quickly. Read the error message carefully to understand what went wrong.

​
Testing Best Practices

​
Before Testing

1

Verify Configuration

  • All API keys are valid and have proper permissions
  • Prompts are complete and reference the correct fields
  • Agent settings match your intended use case
2

Check Connections

  • All agents are properly connected
  • Data flows in the correct direction
  • No circular connections exist
3

Save Everything

  • Save all agent configurations
  • Save the complete workflow
  • Wait for confirmation messages

​
During Testing

  • Test incrementally - Test each new agent before adding more
  • Start simple - Begin with 2-3 agents, then expand
  • Monitor output - Check that each agent produces expected results
  • Note execution time - Track how long the workflow takes
  • Watch for errors - Pay attention to error messages and warnings

​
After Testing

  • Review all outputs - Verify each agent produced correct results
  • Check external services - Confirm emails were sent, messages posted, etc.
  • Analyze performance - Identify slow agents or bottlenecks
  • Document findings - Note any issues or unexpected behavior
  • Iterate and improve - Adjust configurations based on test results

​
Common Testing Scenarios

​
Testing Text Generation

Agent: Text Generator
Input: Email body, sender, subject
Prompt: "Summarize this email: {{email.body}}"
Expected: Brief summary of email content
Test individually first, then test as part of larger workflow.

​
Testing Email Workflows

Workflow: Gmail Reader → Text Generator → Gmail Sender
1. Test Gmail Reader - Verify emails are retrieved
2. Test Text Generator - Verify summaries are created
3. Test Gmail Sender - Verify emails are sent
4. Test complete workflow - Verify end-to-end flow

​
Testing Multi-Output Workflows

Workflow: GitHub Reader → Text Generator → Discord + Email
1. Test GitHub Reader - Verify commits are read
2. Test Text Generator - Verify reports are generated
3. Test Discord Messenger - Verify messages are posted
4. Test Gmail Sender - Verify emails are sent
5. Test complete workflow - Verify all outputs are delivered

​
Debugging Failed Tests

​
Agent Test Failures

1

Check Error Message

Read the error message in the red box. It usually indicates the specific problem.
2

Verify API Keys

  • Ensure API keys are correct
  • Check that keys have necessary permissions
  • Verify keys haven’t expired
3

Review Configuration

  • Check all required fields are filled
  • Verify field references are correct
  • Ensure model selection is appropriate
4

Test Simpler Version

  • Remove field references and use static text
  • Simplify the prompt or configuration
  • Test with minimal settings

​
Workflow Test Failures

1

Identify Failed Agent

Check the error message to see which agent failed.
2

Test That Agent Individually

Open the failed agent’s configuration and test it in isolation.
3

Check Data Flow

Verify the agent is receiving expected input from previous agents.
4

Verify Connections

Ensure all connections are properly established and pointing to the right agents.
5

Test Incrementally

Test partial workflows (first 2 agents, then 3, etc.) to isolate the issue.

​
Testing with Different Data

​
Using Sample Data

For agents that receive input:
  • Start with simple, predictable input
  • Test with minimal data first
  • Gradually increase complexity
  • Test edge cases (empty data, special characters, etc.)

​
Testing External Services

Gmail Agents:
  • Use a test Gmail account
  • Send test emails to yourself
  • Check spam/trash folders
Discord Agents:
  • Create a test Discord server
  • Use a dedicated test channel
  • Monitor for rate limits
GitHub Agents:
  • Use a test repository
  • Create test commits
  • Verify webhook configurations

​
Performance Testing

​
Measuring Execution Time

Track how long your workflow takes:
  1. Note the start time when clicking “Test Workflow”
  2. Note the end time when receiving success message
  3. Calculate total execution time
  4. Identify slow agents
Most workflows should complete in under 30 seconds. If yours takes longer, consider optimizing agent configurations or splitting into multiple workflows.

​
Optimization Tips

  • Use faster AI models for simple tasks
  • Minimize unnecessary text processing
  • Reduce the number of API calls
  • Cache results when possible
  • Test during off-peak hours to avoid rate limits

​
Next Steps

Build docs developers (and LLMs) love

Get started for freeTalk to us