MCP Tool

The MCP Tool node allows you to add individual Model Context Protocol (MCP) tools to your workflows. These tools extend the capabilities of your agents by providing specialized functions they can invoke, such as web search, file operations, API calls, data extraction, and more. MCP tools follow a standardized protocol that enables agents to discover and use them intelligently.

Why Use MCP Tool?

The MCP Tool is ideal when you need to:

Extend agent capabilities - Add specific functions that agents can use to accomplish tasks
Integrate external services - Connect to APIs and external systems through standardized tools
Provide specialized functions - Give agents access to domain-specific operations
Enable tool usage - Allow agents to perform actions beyond language model capabilities
Build custom integrations - Create workflows that combine AI reasoning with external tools
Grant selective access - Choose specific tools for specific agents rather than entire server collections

How It Works

MCP Tools provide a standardized interface for agents to interact with external functions:

Tool Registration - Tools are registered and made available to connected agents
Agent Invocation - Agents can call tools based on their prompts and reasoning
Parameter Passing - Agents provide appropriate parameters when invoking tools
Result Processing - Tool outputs are returned to agents for further processing
Context Awareness - Tools understand the context of the workflow and execution

When to Use MCP Tool

Use MCP Tool when:

You need to add a specific, individual tool to your workflow
The tool provides a single, focused function
You want granular control over which tools are available to which agents
You're building workflows with selective tool access for different agents
You need one or a few specific tools from a server (use MCP Server for full collections)

Configuration Steps

Setting up an MCP Tool involves selecting the server source and choosing the specific tool you want to make available.

Step 1: Server Configuration

First, choose where your MCP tool will come from - either internal pre-configured servers or external custom servers.

Server Type:

Choose between two server types:

Internal Server (Recommended) - Use pre-configured internal MCP servers
- Ready-to-use servers provided by FlowGenX platform
- No additional configuration required
- Pre-tested and verified tools
- Managed authentication and credentials
- Ideal for most use cases
External Server - Connect to custom external MCP servers
- Connect to your own MCP server implementations
- Use third-party MCP servers
- Requires server URL and authentication configuration
- Useful for custom integrations and specialized tools

Step 2: Select MCP Server

After choosing the server type, select the specific MCP server that contains the tool you want to use.

Selecting from Internal Servers:

Click on the Select MCP Server dropdown
Browse available internal MCP servers
Each server displays:
- Server name and identifier
- Description of the server's purpose and capabilities
- Icon or logo for easy identification

Example Server:

scrapy-jane-doe-20260107-115640
- Description: "Web scraping connector powered by Scrapy framework for extracting data from websites using powerful CSS selectors..."
- Provides tools for web scraping and content extraction

Once selected, the server information is displayed with its description, helping you understand what tools it provides.

Step 3: Select MCP Tool

After selecting a server, choose the specific tool from that server you want to configure.

Click on the Select MCP Tool dropdown
Browse available tools from the selected server
View tool names and brief descriptions
Select the tool that matches your workflow needs

Example Tool:

Get full page content - Retrieves complete page content including HTML, text, title, and metadata

Step 4: Tool Summary

After selecting a tool, the Tool Summary section displays comprehensive information about the configured tool:

Tool Summary Information:

Tool Name - The name of the selected tool
- Example: "Get full page content"
Server - The MCP server providing this tool
- Example: "scrapy-jane-doe-20260107-115640"
Description - Detailed explanation of what the tool does
- Example: "Retrieves complete page content including full HTML, plain text, title, and metadata. Perfect for cases where you want the whole page without specifying selectors. Useful for content archiving, AI processing, or when you need to analyze the full page before deciding what to extract."

This summary helps you verify that you've selected the correct tool and understand exactly what it will do when invoked by an agent.

Connecting to Agents

Once configured, the MCP Tool node can be connected to agents in your workflow to make the tool available for their use.

Connecting to React Agent

To connect the MCP Tool to a React Agent:

Add both MCP Tool node and React Agent node to your workflow
Draw a connection from the MCP Tool to the React Agent
The React Agent will automatically detect the available tool
Update the React Agent's prompt to explain when and how to use the tool
The agent can now invoke the tool during execution

Benefits for React Agent:

Single agent gains access to the specific tool
Agent can use the tool as part of its task execution
Tool results enhance the agent's capabilities
Clear, focused tool availability

For detailed information on configuring React Agents, see the React Agent documentation.

Connecting to Supervisor Agent

To connect the MCP Tool to a Supervisor Agent:

Add both MCP Tool node and Supervisor Agent node to your workflow
Draw a connection from the MCP Tool to the Supervisor Agent
The Supervisor Agent will detect the tool and can delegate its use
Update the Supervisor's prompt to explain the tool's purpose and when to use it
The supervisor can invoke the tool directly or instruct child agents to use it

Benefits for Supervisor Agent:

Supervisor can coordinate tool usage across multiple agents
Tool can be used as part of complex orchestration strategies
Supervisor can decide when and how to invoke the tool
Enables intelligent tool delegation based on task requirements

For detailed information on configuring Supervisor Agents, see the Supervisor Agent documentation.

Using the MCP Tool in Your Workflow

After configuration and connection:

The MCP Tool node appears in your workflow canvas
Agents connected to the tool can discover and invoke it
When an agent needs the tool's functionality, it can call it with appropriate parameters
The tool executes its function (e.g., web scraping, API call, data processing)
Results are returned to the agent for use in generating responses or making decisions
The workflow continues with the enriched data from the tool

Tool Invocation Flow:

Agent receives task → Determines tool is needed →
Invokes MCP Tool with parameters →
Tool executes function →
Returns results to agent →
Agent processes results → Continues workflow

Best Practices

Tool Selection

Choose specific tools - Select only the tools you actually need for your workflow
Understand tool capabilities - Read the tool description carefully before selection
Match tools to tasks - Ensure the tool's function aligns with your workflow requirements
Avoid redundancy - Don't add multiple tools that do the same thing

Server Selection

Prefer internal servers - Use internal servers when available for easier management
Verify server reliability - Ensure external servers are stable and available
Check server documentation - Understand what tools each server provides
Consider server maintenance - Internal servers are maintained by the platform

Agent Integration

Document in prompts - Clearly explain the tool's purpose in your agent prompts
Provide usage examples - Show agents when and how to use the tool
Set clear boundaries - Define when the tool should and shouldn't be used
Test thoroughly - Verify tool invocation works correctly before deployment

Tool Configuration

Validate selections - Check the Tool Summary to confirm correct tool selection
Review descriptions - Ensure you understand what the tool will do
Test tool output - Verify the tool returns expected results
Monitor usage - Track how often and how successfully the tool is being used

Common Use Cases

Web Scraping and Data Extraction:

Connect web scraping tools to agents for content retrieval
Extract structured data from websites
Monitor web pages for changes
Gather information for analysis or reporting

API Integration:

Give agents access to external API calls
Retrieve data from third-party services
Send data to external systems
Integrate with business applications

File Operations:

Enable agents to read and write files
Process documents and data files
Generate reports and exports
Manage file systems

Data Processing:

Transform and enrich data
Validate and clean datasets
Perform calculations and analysis
Format data for downstream use

Search and Retrieval:

Search databases and knowledge bases
Query external search engines
Retrieve relevant information
Find specific data points

Troubleshooting

Tool Not Available to Agent:

Verify the MCP Tool node is connected to the agent node
Check that the tool is properly configured
Ensure the server is accessible and active
Review agent logs for tool discovery issues

Tool Invocation Failing:

Validate tool parameters are correct
Check server connectivity and status
Verify authentication and permissions
Review tool-specific error messages

Unexpected Tool Results:

Review the tool description to confirm expected behavior
Check input parameters sent by the agent
Validate the data being processed by the tool
Test the tool independently to verify functionality

Agent Not Using Tool:

Ensure the tool is mentioned in the agent's prompt
Provide clear instructions on when to use the tool
Add examples of tool usage to the prompt
Check if the agent has other ways to accomplish the task

Monitoring Tool Usage

Monitor your MCP Tool performance:

Track invocation frequency - How often agents are using the tool
Measure success rates - Percentage of successful tool calls
Monitor response times - How long tools take to execute
Review tool outputs - Verify tools are returning useful results
Analyze agent behavior - Understand when and why agents invoke tools

For more information on monitoring workflow executions, see the Traceability documentation.

MCP Tool vs MCP Server

Use MCP Tool when:

You need one or a few specific tools
You want granular control over tool access
Different agents need different tools
You're building focused, specialized workflows

Use MCP Server when:

You need multiple tools from the same server
You want to provide comprehensive toolsets
All tools from a server are relevant to your workflow
You prefer centralized tool management

For information on using MCP Servers, see the MCP Server documentation.

Next Steps

After setting up your MCP Tool:

Connect the tool to your React Agent or Supervisor Agent
Update agent prompts to explain the tool's purpose and usage
Test tool invocation in the Playground
Verify tool results are as expected
Deploy your workflow with the integrated tool
Monitor tool usage and performance

MCP Tools bring powerful external capabilities to your FlowGenX workflows, enabling agents to perform specialized tasks and integrate with external systems beyond their native language model capabilities.