Skip to main content
Execute and manage your AI agents with confidence. Learn how to run agents manually, set up automated execution, monitor performance, and troubleshoot issues for reliable automation workflows.

​
Overview

Running agents in mixus involves multiple execution methods, from manual one-time runs to sophisticated automated workflows. Whether you need immediate task execution or ongoing automation, mixus provides flexible options to meet your requirements.

​
How Agent Execution Works

Agent execution follows a structured process:
  1. Trigger Activation: Agent receives a trigger (manual, scheduled, or event-based)
  2. Context Loading: Agent loads its instructions, integrations, and available tools
  3. Task Processing: Agent executes the requested task using AI reasoning and tools
  4. Integration Actions: Agent interacts with connected services as needed
  5. Output Generation: Agent produces results, notifications, or follow-up actions
  6. Logging & Monitoring: Execution details are logged for review and optimization

Agent execution flow

Image placeholder - To be added

​
Execution Methods

​
Manual Execution

​
Direct Agent Activation

Run agents immediately for specific tasks: 
# In any chat, mention your agent
@customer-support-agent "Handle this customer inquiry about billing"

# Direct agent conversation
Start a conversation with: Customer Support Agent

# Dashboard execution
Navigate to Agents → [Agent Name] → Run Now

​
Chat-Based Execution

Integrate agent capabilities into ongoing conversations: 
# Request agent assistance in team chat
"@sales-agent Generate a proposal for the client we discussed"

# Use agent in analysis workflow
"@data-agent Analyze the attached spreadsheet and create visualizations"

# Multi-agent coordination
"@research-agent Find market data, then @report-agent create a summary"

​
Natural Language Commands

You can also run agents using simple natural language commands: 
# Basic syntax
"run agent {agent name}"

# Examples
"run agent customer support"
"run agent daily report generator"
"execute agent sales assistant"

# With additional context
"run agent data processor with this CSV file"
"run agent email responder but focus on urgent messages"

# When multiple agents have similar names
"run agent marketing" → System will show all matching agents
Select specific agent: "run agent 1" or use the specific agent ID
How it works:
  • The system recognizes “run agent” or “execute agent” followed by the agent name
  • If multiple agents match the name, you’ll see a list with details (name, ID, creator, description)
  • You can add context with “with this data” or modify behavior with “but…” instructions
  • The agent runs immediately in the current chat context

​
Scheduled Execution

​
Time-Based Scheduling

Set up agents to run automatically at specific times:
Schedule Configuration
{
  "schedule_type": "recurring",
  "frequency": "daily",
  "time": "09:00",
  "timezone": "America/New_York",
  "days": ["monday", "tuesday", "wednesday", "thursday", "friday"],
  "enabled": true
}

​
Advanced Scheduling Options

Advanced Scheduling
{
  "schedule_type": "cron",
  "cron_expression": "0 0 1 * *",  // First day of every month
  "timezone": "UTC",
  "retry_policy": {
    "max_attempts": 3,
    "delay_minutes": 15
  },
  "notifications": {
    "on_success": ["team-lead@company.com"],
    "on_failure": ["admin@company.com", "#alerts"]
  }
}

​
Event-Based Execution

​
Webhook Triggers

Configure agents to respond to external events:
Webhook Configuration
{
  "trigger_type": "webhook",
  "webhook_url": "https://api.mixus.ai/webhooks/agent-123",
  "authentication": {
    "type": "signature",
    "secret": "your-webhook-secret"
  },
  "filters": {
    "event_type": "new_lead",
    "source": "website_form",
    "priority": ["high", "urgent"]
  }
}

​
Integration Triggers

Set up agents to respond to changes in connected services: 
# Email-based triggers
"Run when new emails arrive in support inbox"

# CRM triggers
"Execute when lead status changes to 'qualified'"

# Calendar triggers
"Activate 15 minutes before scheduled meetings"

# Slack triggers
"Run when mentioned in #emergency channel"

​
Execution Contexts

​
Individual Agent Runs

​
Single-Task Execution

Perfect for specific, one-time tasks: 
# Data analysis task
Agent: "Analyze last quarter's sales data and identify top performers"

# Customer communication
Agent: "Send follow-up emails to all trial users ending this week"

# Content creation
Agent: "Create social media posts for next week's product launch"

​
Parametric Execution

Pass specific parameters for customized execution:
Execution Parameters
{
  "agent_id": "sales-report-agent",
  "parameters": {
    "date_range": "2024-01-01 to 2024-03-31",
    "departments": ["sales", "marketing"],
    "format": "executive_summary",
    "recipients": ["ceo@company.com", "sales@company.com"]
  }
}

​
Multi-Agent Workflows

​
Sequential Execution

Chain multiple agents for complex workflows: 
1. Research Agent: "Gather market data on competitors"
2. Analysis Agent: "Analyze the research data for trends"
3. Report Agent: "Create executive summary"
4. Distribution Agent: "Send report to stakeholder list"

​
Parallel Execution

Run multiple agents simultaneously: 
Trigger simultaneously:
- Social Media Agent: "Create content for professional networks"
- Email Agent: "Draft newsletter for customers"
- Analytics Agent: "Generate weekly performance report"
- Meeting Agent: "Schedule follow-up meetings"

​
Monitoring and Management

​
Execution Dashboard

​
Real-Time Monitoring

Track agent performance through the dashboard: 
✓ Active Executions: 3 agents currently running
✓ Success Rate: 94.2% (last 30 days)
✓ Average Runtime: 2.3 minutes
✓ Queue Status: 2 agents queued, 0 errors
✓ Integration Health: All systems operational

​
Historical Analysis

Review past executions for optimization: 
📊 Execution History (Last 7 Days):
- Total Runs: 156
- Successful: 148 (94.9%)
- Failed: 5 (3.2%)
- Cancelled: 3 (1.9%)
- Average Duration: 1.8 minutes

​
Live Execution Tracking

​
Real-Time Status Updates

Monitor agents as they work: 
🔄 Sales Report Agent - Running (2:34 elapsed)
├── ✅ Connected to CRM system
├── 🔄 Analyzing Q1 sales data...
├── ⏳ Generating revenue charts
└── ⏳ Preparing executive summary

Estimated completion: 3-5 minutes

​
Step-by-Step Progress

See detailed execution steps: 
Customer Support Agent - Processing Inquiry

Step 1: ✅ Analyzed customer message (0:15)
Step 2: ✅ Retrieved customer history from CRM (0:23)
Step 3: ✅ Identified issue category: billing (0:05)
Step 4: 🔄 Searching knowledge base for solution... (0:45)
Step 5: ⏳ Preparing personalized response
Step 6: ⏳ Scheduling follow-up if needed

​
Performance Optimization

​
Execution Time Analysis

Identify bottlenecks and optimization opportunities:
Performance Breakdown
{
  "agent": "email-automation-agent",
  "average_execution_time": "3.2 minutes",
  "breakdown": {
    "initialization": "0.3 minutes",
    "data_retrieval": "1.8 minutes",
    "processing": "0.7 minutes",
    "integration_calls": "0.4 minutes"
  },
  "optimization_suggestions": [
    "Cache frequently accessed data",
    "Batch integration API calls",
    "Optimize search queries"
  ]
}

​
Resource Usage Monitoring

Track computational and integration resources: 
📈 Resource Utilization (Last Hour):
- CPU Usage: 23% average, 67% peak
- Memory: 1.2GB average, 2.1GB peak
- API Calls: 156 total, 2.6/minute average
- Integration Quotas: 67% of daily limits used

​
Error Handling and Recovery

​
Common Execution Issues

​
Integration Failures

Handle service connectivity problems: 
❌ Error: CRM connection timeout
🔄 Auto-retry: Attempt 2 of 3 in 30 seconds
📝 Fallback: Using cached customer data
✅ Continued execution with degraded functionality
📧 Alert sent to integration admin

​
Authentication Errors

Manage credential and permission issues: 
❌ Email integration: Token expired
🔧 Auto-action: Refreshing OAuth token
✅ Token refreshed successfully
▶️ Resuming email processing
📝 Log: Authentication renewed at 14:23

​
Rate Limiting

Handle API rate limit constraints: 
⚠️ Warning: Approaching social media API rate limit
📊 Current usage: 148/180 requests (next reset: 15 minutes)
🔄 Implementing exponential backoff
⏸️ Queuing remaining posts for next window
📈 Estimated completion: 18 minutes

​
Automatic Recovery

​
Retry Mechanisms

Configure intelligent retry strategies:
Retry Configuration
{
  "retry_policy": {
    "max_attempts": 3,
    "backoff_strategy": "exponential",
    "initial_delay": "30 seconds",
    "max_delay": "5 minutes",
    "retry_conditions": [
      "network_timeout",
      "rate_limit_exceeded",
      "temporary_service_unavailable"
    ]
  }
}

​
Graceful Degradation

Continue operation with reduced functionality: 
🔧 CRM integration unavailable
📋 Switching to local customer database
⚠️ Note: Real-time data may be limited
▶️ Continuing with available information
📧 Customer service team notified

​
Manual Recovery

​
Execution Restart

Resume failed executions from specific points: 
❌ Agent failed at Step 3: Generate report
🔧 Recovery options:
   1. Restart from beginning
   2. Resume from Step 3
   3. Skip Step 3 and continue
   4. Restart with different parameters

​
Emergency Controls

Stop or modify running agents when needed: 
🛑 Emergency stop: Immediately halt all agent operations
⏸️ Pause execution: Temporarily suspend agent
🔧 Modify parameters: Adjust settings during execution
🔄 Force retry: Override automatic retry limits
📝 Emergency log: Document incident and actions taken

​
Best Practices

​
Execution Planning

  1. Start Small and Scale

​
Begin with simple, low-risk tasks

Phase 1: Manual execution with supervision Phase 2: Scheduled execution with monitoring Phase 3: Event-driven execution with full automation Phase 4: Multi-agent workflows with error handling 

2. **Test Before Production**
   ```plaintext
# Development environment testing
   - Test with sample data
   - Verify all integration connections
   - Check error handling scenarios
   - Validate output quality
   
   # Staging environment validation
   - Run with production-like data
   - Test performance under load
   - Verify monitoring and alerting
   - Confirm backup and recovery procedures
  1. Monitor and Optimize

​
Regular performance review

  • Weekly execution reports
  • Monthly optimization sessions
  • Quarterly architecture review
  • Annual strategy assessment

​
Continuous improvement

  • Identify bottlenecks and inefficiencies
  • Update agent instructions based on results
  • Optimize integration usage patterns
  • Enhance error handling based on common issues

### Security and Compliance

1. **Access Control**
   - Use least-privilege principle for agent permissions
   - Regularly audit and rotate integration credentials
   - Implement approval workflows for sensitive operations
   - Monitor agent actions for security compliance

2. **Data Protection**
   - Encrypt sensitive data in agent configurations
   - Implement data retention policies for execution logs
   - Ensure compliance with privacy regulations
   - Use secure channels for all external communications

3. **Audit and Logging**
   - Maintain detailed logs of all agent executions
   - Implement tamper-proof audit trails
   - Regular compliance reporting and review
   - Document all changes to agent configurations

## Troubleshooting

### Performance Issues

#### Slow Execution
**Problem**: Agent takes longer than expected to complete tasks  
**Solutions**:
- Review integration response times
- Optimize data queries and processing logic
- Implement caching for frequently accessed data
- Consider breaking complex tasks into smaller steps

#### High Resource Usage
**Problem**: Agent consumes excessive CPU or memory  
**Solutions**:
- Analyze processing logic for inefficiencies
- Implement pagination for large data sets
- Optimize file processing and memory usage
- Scale infrastructure if consistently high usage is justified

#### Frequent Timeouts
**Problem**: Agent executions frequently timeout  
**Solutions**:
- Increase timeout limits for complex operations
- Implement progress checkpoints for long-running tasks
- Optimize external API call patterns
- Consider asynchronous processing for time-intensive operations

### Reliability Issues

#### Inconsistent Results
**Problem**: Agent produces different outputs for similar inputs  
**Solutions**:
- Review and clarify agent instructions
- Implement input validation and normalization
- Add deterministic processing steps where possible
- Monitor for external data source variations

#### Integration Failures
**Problem**: External service connections frequently fail  
**Solutions**:
- Implement robust retry mechanisms
- Add circuit breaker patterns for unreliable services
- Create fallback data sources or processes
- Monitor service status and plan for outages

## Related Features

- [Creating Agents](/agents/creating) - Learn how to build effective agents
- [Agent Scheduling](/agents/scheduling) - Set up automated execution schedules
- [Integrations](/integrations/overview) - Connect agents to external services
- [Monitoring & Analytics](/monitoring/overview) - Track agent performance

## What's Next?

Ready to start running your agents effectively? Here are your next steps:

1. **[Create your first agent](/agents/creating)** if you haven't already
2. **[Set up monitoring](/monitoring/agents)** to track performance
3. **[Configure scheduling](/agents/scheduling)** for automated execution
4. **[Explore agent examples](/agents/examples)** for inspiration and best practices

---

*Need help optimizing your agent execution? Contact our [agent specialists](mailto:support@mixus.com) or join our [community](/support/community) for tips and best practices.*
⌘I