Agentic Workers
Core featuresExecution modes

Agent Mode

Execute workflows directly on Agentic Workers with ultra-fast inference

Agent Mode

Agent Mode

Agent Mode executes workflows directly on Agentic Workers, providing ultra-fast inference and dedicated workflow management. It's optimized for complex, multi-step tasks and background execution.

What is Agent Mode?

Agent Mode is Agentic Workers' native execution environment. When you run a workflow in Agent Mode:

  • Ultra-fast inference: Optimized for speed and performance
  • Dedicated interface: Purpose-built for workflow execution
  • Background tasks: Run long workflows without blocking
  • Full tool access: All integrations available
  • Activity tracking: Monitor progress in real-time

When to Use Agent Mode

Use Agent Mode When:

  • You need the fastest possible execution
  • Running complex, multi-step workflows
  • Executing long-running tasks (background mode)
  • You want a dedicated workflow interface
  • You need advanced features like prompt chaining
  • You're using many tool integrations

Use Chat Mode When:

  • You're already working in ChatGPT, Claude, or Gemini
  • You want to combine workflows with regular chat
  • You prefer the familiar AI platform interface

Running Workflows in Agent Mode

Method 1: From the Dashboard

  1. Navigate to Prompt Library
  2. Select a template
  3. Fill in variables (if required)
  4. Click "Start Workflow"
  5. Choose "Agent Mode" if prompted
  6. Monitor progress in the activity drawer

Method 2: From Agent Chat

  1. Open an agent from the Agents section
  2. Start a new chat session
  3. Enter your workflow prompt
  4. The workflow executes in Agent Mode automatically
  5. View results in the chat interface

Method 3: From Your Library

  1. Go to My Library or Prompt Library
  2. Click on a saved template
  3. Click "Run" or "Start Workflow"
  4. Select Agent Mode execution
  5. Monitor progress

Background Tasks

Agent Mode supports background execution for long-running workflows:

Starting a Background Task

  1. Start a workflow as usual
  2. If the workflow is long-running, you'll see an option to run in background
  3. Click "Run in Background"
  4. Continue working on other tasks
  5. Check the Activity drawer for progress

Monitoring Background Tasks

  1. Open the Activity drawer (sidebar icon)
  2. View all running and completed workflows
  3. Click on any task to see details
  4. View results when complete

Background Task Benefits

  • Non-blocking: Continue working while workflows run
  • Progress tracking: See real-time updates
  • Result storage: Access results later
  • Multiple tasks: Run several workflows simultaneously

Agent Mode Features

Ultra-Fast Inference

Agent Mode is optimized for speed:

  • Direct API connections
  • Optimized model routing
  • Reduced latency
  • Parallel execution when possible

Full Tool Access

All your connected integrations are available:

  • Google Workspace (Docs, Sheets, Calendar)
  • Communication (Slack, Gmail, Outlook)
  • Productivity (Notion, Airtable, Linear)
  • Research (Perplexity, Reddit, Firecrawl)
  • And many more...

Prompt Chaining

Agent Mode fully supports prompt chaining:

  • Multi-step workflows
  • Data passing between steps
  • Conditional execution
  • Complex automation

Activity Tracking

Monitor all Agent Mode executions:

  • Real-time progress updates
  • Step-by-step results
  • Error handling and retries
  • Cost and token tracking

Differences from Chat Mode

FeatureAgent ModeChat Mode
SpeedUltra-fastPlatform-dependent
InterfaceDedicated workflow UIAI platform chat
Background Tasks✅ Supported❌ Not supported
Tool Access✅ Full access✅ Full access
Prompt Chaining✅ Advanced✅ Basic
Activity Tracking✅ Detailed⚠️ Limited

Performance Tips

Optimize Workflow Design

  • Break complex workflows into steps
  • Use variables efficiently
  • Minimize redundant operations
  • Cache results when possible

Model Selection

Choose the right model for your task:

  • GPT-4o-mini: Fast, cost-effective for simple tasks
  • GPT-4o: Best for complex reasoning
  • Claude 3.5 Sonnet: Excellent for long-form content
  • Gemini Pro: Great for multimodal tasks

Tool Usage

  • Connect only needed integrations
  • Use tools efficiently
  • Avoid redundant API calls
  • Batch operations when possible

Troubleshooting

Slow Execution

  • Check model selection (use faster models for simple tasks)
  • Review workflow complexity
  • Check integration connection status
  • Monitor activity drawer for errors

Workflow Errors

  • Review error messages in activity drawer
  • Check variable values
  • Verify tool connections
  • Test workflow step-by-step

Background Task Issues

  • Check activity drawer for status
  • Review error logs
  • Verify workflow syntax
  • Retry failed steps if needed

Next Steps