Skip to main content

​
Tools Overview

Tools extend Xagent’s capabilities by enabling agents to perform actions beyond text generation. Instead of just responding, agents can search the web, process files, analyze images, execute code, and more.

​
What Are Tools?

Tools are functions that agents can call during execution to perform specific actions:
  • Web Search - Find current information online
  • Knowledge Base - Search your documents
  • File Operations - Read, write, and process files
  • Vision - Understand and analyze images
  • Image Generation - Create and edit images
  • Code Execution - Run Python code
  • Browser Automation - Control web browsers
  • And more - Extensible through MCP protocol

​
Why Use Tools?

Tools enable agents to:
  • Take real-world actions - Interact with actual systems
  • Access dynamic data - Get up-to-date information
  • Process files - Work with documents, images, code
  • Automate workflows - Execute multi-step operations

​
Tool Categories

Xagent organizes tools into categories for easy management:

​
Basic Tools

Core capabilities for general tasks:
  • Web Search - Search the internet for current information
  • Python Executor - Execute Python code for data analysis and computation
Use for: General-purpose tasks that need web access or computation

​
Knowledge Tools

Search and retrieve information from your knowledge bases:
  • List Knowledge Bases - See available knowledge bases
  • Search Knowledge Base - Find relevant documents using RAG
Use for: Domain-specific questions, documentation, FAQs

​
File Tools

Work with files in your workspace:
  • Read Files - Read text files, documents, PDFs
  • Write Files - Save content to files
  • List Files - Browse directories
  • Edit Files - Modify file content
  • Parse Documents - Extract text from PDFs, DOCX, Excel, etc.
Use for: Document processing, data analysis, file management

​
Vision Tools

Understand and analyze visual content:
  • Understand Images - Answer questions about images
  • Describe Images - Generate detailed descriptions
  • Detect Objects - Find and identify objects in images
Use for: Image analysis, OCR, chart understanding, visual data extraction Requires: Vision model (multimodal LLM)

​
Image Tools

Create and modify images:
  • Generate Image - Create images from text descriptions
  • Edit Image - Modify existing images
Use for: Content creation, design, marketing materials Requires: Image generation model (DALL-E, Stable Diffusion, etc.)

​
Browser Tools

Automate web browser interactions:
  • Navigate - Browse websites
  • Interact - Click buttons, fill forms
  • Extract - Scrape web data
  • Screenshot - Capture pages
Use for: Web scraping, automation, testing, data collection

​
MCP Tools

Connect to external tools via MCP (Model Context Protocol):
  • Dynamic tool loading
  • External service integration
  • Custom tool servers
Use for: Integrating with specialized tools and services

​
Agent Tools

Use published agents as tools:
  • Task delegation
  • Specialized agent capabilities
  • Workflow composition
Use for: Complex multi-agent workflows

​
Using Tools in Agents

​
Configuring Tool Access

When building an agent, you control which tools it can use:
  1. Go to Build page
  2. Create or edit an agent
  3. In the Tools section:
    • Select tool categories to enable
    • Agent will only have access to tools in chosen categories
    • Different agent roles need different tools
Examples:
  • Customer Support Agent - Enable: knowledge, file
  • Data Analyst Agent - Enable: basic, file, python executor
  • Content Creator Agent - Enable: basic, image, file
  • Research Assistant - Enable: basic, knowledge, browser, file

​
Tool Categories in Agent Builder

Available categories to select:
  • basic - Web search, code execution
  • file - File operations and document parsing
  • vision - Image understanding
  • image - Image generation and editing
  • knowledge - Knowledge base search
  • browser - Browser automation
  • ppt - PowerPoint processing
  • office - Office document processing
  • special_image - Special image operations
  • mcp - MCP protocol tools
Only enable tool categories that the agent needs. Too many tools can lead to slower execution and irrelevant tool calls.

​
How Tools Work

​
Tool Selection

When executing a task, Xagent:
  1. Analyzes the request - Understands what needs to be done
  2. Selects relevant tools - Chooses appropriate tools from available categories
  3. Calls the tool - Executes with required parameters
  4. Processes results - Uses tool output to generate response

​
Tool Parameters

Each tool has specific parameters:
  • Required parameters - Must be provided
  • Optional parameters - Have default values
  • Auto-generated parameters - Filled in by Xagent (like file paths)

​
Tool Results

Tools return structured results:
  • Success status - Whether the tool call succeeded
  • Output data - Result of the operation
  • Error messages - Details if something went wrong
  • Metadata - Additional information (execution time, tokens, etc.)

​
Best Practices

​
For Agent Builders

Match Tools to Purpose
  • Only enable necessary tool categories
  • Consider what actions the agent needs to perform
  • Avoid over-provisioning tools
Test Tool Access
  • Verify agent can use enabled tools
  • Test with realistic tasks
  • Monitor tool usage patterns
Consider Security
  • Some tools have security implications (file, browser, code)
  • Only enable for trusted use cases
  • Review agent permissions

​
For Users

Be Specific in Requests
  • Clear requests help Xagent choose the right tools
  • Mention what you want to accomplish
  • Provide context about available resources
Trust the Process
  • Xagent automatically selects appropriate tools
  • You don’t need to specify which tool to use
  • Focus on describing what you want
Provide Feedback
  • If Xagent uses the wrong tool, clarify your request
  • Suggest the specific action you want
  • Correct misunderstandings

​
Next Steps