Skip to main content

Agents Overview

Agents are reusable AI assistants with predefined roles and configurations. Unlike one-time tasks, agents maintain consistent behavior and can be shared across your team.

Agent vs Task

When to Use Tasks

Tasks are best for:
  • One-time executions
  • Quick exploratory work
  • Testing ideas
  • Ad-hoc requests
  • Temporary workflows
Example: “Summarize this document” - a one-time request that doesn’t need to be saved or reused.

When to Use Agents

Agents are best for:
  • Reusable workflows
  • Repeated operations
  • Team collaboration
  • Production use cases
  • Specialized roles
Example: A “Customer Support Agent” that handles FAQ requests, searches documentation, and provides consistent responses.

Comparison

FeatureTasksAgents
Use caseOne-time executionsReusable workflows
ConfigurationAuto-generatedCustomizable
PersistenceNot savedSaved and shareable
BehaviorVaries by requestConsistent
SharingNot shareableCan be published
Role definitionPer taskPredefined

Key Agent Features

1. Role Definition

Define your agent’s purpose and behavior:
  • “You are a customer support assistant”
  • “You are a data analyst”
  • “You are a research assistant”

2. Execution Mode

Choose how your agent processes tasks:
  • React Mode: Reasoning → Acting cycle (default)
  • Graph Mode: Plan all steps first, then execute

3. Tool Access

Control what tools your agent can use:
  • Basic tools (web search, calculator)
  • File operations
  • Image understanding/generation
  • Knowledge base operations
  • And more

4. Knowledge Integration

Add domain-specific knowledge:
  • Attach knowledge bases
  • Agent retrieves relevant information
  • Enhances response accuracy

5. Model Configuration

Select models for different purposes:
  • General Model: Main reasoning engine
  • Fast Model: For simple operations
  • Visual Model: For image tasks
  • Compact Model: For long conversations

How It Works

Creating Tasks

For one-time tasks, simply:
  1. Go to Task page
  2. Enter your request
  3. Upload files if needed
  4. Xagent executes and shows results

Creating Agents

For reusable workflows:
  1. Go to Build page
  2. Create new agent
  3. Configure role, models, tools, knowledge
  4. Test with live preview
  5. Save and publish
After publishing, users can:
  • Find the agent in the agents list
  • Start a conversation with the agent
  • Agent uses its configured behavior consistently

Shared Execution Interface

Both tasks and agents use the same execution interface:
  • Real-time progress tracking
  • Step-by-step execution details
  • Tool calls and responses
  • Chat interaction
When using an agent, the interface displays:
  • Agent’s logo and name
  • Agent’s configuration
  • Consistent behavior based on agent’s settings

Typical Workflow

For One-Time Tasks

Task Page → Enter Request → Execution → Results
Done! No need to save or configure anything.

For Creating Agents

Build Page → Create Agent → Configure → Test → Save → Publish
After publishing:
Agents List → Select Agent → Chat → Agent responds with configured behavior

Agent Templates

Jump-start agent creation with pre-built templates:
  • Customer Support: FAQ handling, documentation search
  • Data Analyst: Data processing, visualization, reporting
  • Research Assistant: Information gathering, synthesis
  • Content Creator: Writing, editing, image generation
Templates provide a starting point that you can customize for your specific needs. Explore templates →

Next Steps