Who This Is For / When to Use
Use MCP tools when you want your AI Agent to:
Access real-time external data (web, APIs, databases)
Go beyond data stored in your account
Perform live research or enrichment during workflows
Connect to custom or third-party services
What’s New
The AI Agent action now supports MCP connections as tools.
This enables:
Connecting to any MCP-compatible server
Discovering available tools automatically
Enabling or disabling tools per agent
The agent can now interact with external systems directly during workflow execution.
What MCP Tools Enable
MCP tools allow the AI Agent to:
Search the web in real time
Fetch and read web pages
Query external APIs
Access databases or internal tools
Automate browser-based tasks
This expands the agent’s capabilities beyond internal data.
How MCP Connections Work
You connect an MCP-compatible server
The system discovers available tools from that server
You select which tools the AI Agent can use
The agent calls those tools during workflow execution
Each tool acts as an extension of the AI Agent.
How to Add MCP Tools to an AI Agent
Open your workflow
Click the AI Agent action
Click Add Tools
Navigate to the MCP tab
Click Add Connection
Configure the connection:
Connection Name
Server URL
Transport Type
Authentication method
Click Test Connection to load available tools
Select the tools you want the agent to use
Save the action
MCP Connection Configuration Fields
Connection Name
A label to identify the MCP connection.
Server URL
The endpoint of the MCP server (e.g., https://mcp.exa.ai/mcp).
Transport Type
HTTP Streamable
SSE (Server-Sent Events)
Auth Type
Supported authentication methods:
None
Bearer Token
API Key
OAuth2
Custom Header
Key Settings (if using API Key)
Key Name
Key Value
Pass via Headers or Query Parameters
Tool Discovery and Selection
After testing the connection:
The system automatically lists all available tools
You can enable or disable each tool individually
Only selected tools are available to the AI Agent
Where MCP Tools Appear in the AI Agent
Once added, MCP tools appear inside the AI Agent action alongside native tools.
They can be used during execution like any other tool
You control access per agent instance
Supported MCP Servers
Any MCP-compatible server can be used.
Common examples:
Exa (AI-powered web search)
Tavily (web search and extraction)
Browserbase (browser automation)
Custom or internal MCP servers
Why This Matters
Access External Data in Real Time
The AI Agent can retrieve live information instead of relying only on stored data.
Open Integration Model
MCP is an open standard, allowing connections to a wide range of services.
Flexible and Extensible
You can start with prebuilt tools and expand to custom integrations.
Granular Control
You decide exactly which tools each AI Agent can access.
Common Issues and Fixes
Tools not appearing after connection
Click Test Connection to load available tools
Verify the Server URL is correct
Authentication errors
Confirm API key or auth method is configured correctly
Ensure correct header or query parameter selection
Agent not using MCP tools
Ensure the tool is enabled in the AI Agent action
Update the agent instructions to allow tool usage
FAQ
What is MCP?
MCP (Model Context Protocol) is a standard that allows AI systems to connect to external tools and services.
Do I need coding knowledge to use MCP tools?
No. You can connect prebuilt MCP servers without coding.
Can I connect my own internal tools?
Yes. Any MCP-compatible server, including custom or internal systems, can be connected.
Are MCP tools secure?
Yes. You control authentication and which tools each agent can access.
Can I limit which tools an AI Agent uses?
Yes. You can enable or disable individual tools per agent.