MCP Integrations: Connect External Knowledge and Actions

What are MCP Integrations?

MCP (Model Context Protocol) Integrations allow Eloquens AI to connect with external systems and data sources to enhance your workflow capabilities. Think of MCP as a bridge that lets your AI assistant access real-time information from other tools and perform actions in those systems - all automatically as part of your email workflow.

What Happens Behind the Scenes

When you enable MCP Integrations, Eloquens AI gains new abilities:

External Knowledge Access: AI can search and retrieve information from connected systems like Atlassian (Jira/Confluence), databases, or custom applications

Automated Actions: AI can perform tasks in external systems based on email content - like creating Jira tickets, updating records, or triggering workflows

Real-Time Data: Information is fetched in real-time as emails are processed, ensuring responses use the most current data

Intelligent Tool Selection: AI automatically determines which tools to use based on email content and your configuration

Two Types of MCP Capabilities

MCP Integrations provide two distinct capabilities:

Knowledge Tools

What They Do: Search and retrieve information from external systems to enhance AI responses

Example Use Cases:

• Search Confluence documentation to answer technical questions
• Look up Jira tickets to provide project status updates
• Query customer databases for account information
• Access internal knowledge bases for policy details

When Used: Automatically when AI needs information to craft a response

Action Tools

What They Do: Perform tasks in external systems based on email requests

Example Use Cases:

• Create Jira tickets from bug reports
• Update project status in tracking systems
• Log customer requests in CRM systems
• Trigger workflows in connected applications

When Used: When Intelligent Routing determines an action should be taken based on email content

How to Set Up MCP Integrations

Part 1: Create an MCP Connection

Before you can use MCP in workflows, you need to connect to external systems.

Step 1: Navigate to Knowledge Manager

• Go to Knowledge in the side navigation panel
• Click on the Connections tab at the top
• You'll see sections for different connection types

Step 2: Add a New MCP Connection

• In the MCP Integrations section, click Add Connection (or the "+" icon)
• A dialog will open for connection configuration

Step 3: Configure Connection Details

Fill in the connection information:

• Name: Give your connection a descriptive name (e.g., "Atlassian MCP Server", "Customer Database")
• Server URL: Enter the MCP server endpoint URL
  • This is provided by your MCP server (e.g., https://mcp-server.example.com/sse)
• Authentication: Select the authentication method
  • Choose from API Key (Bearer token), API Key (Custom header), or API Key (Query parameter)
• API Key: Enter your authentication credentials
  • For custom header authentication, you'll also specify the header name
  • For query parameter authentication, you'll specify the parameter name

Step 4: Add AI Instructions (Optional)

• In the AI Instructions field, provide guidance for how AI should use this connection
• Examples:
  • "Use this for customer account lookups"
  • "Search this when users ask about technical documentation"
  • "Only create tickets for bugs, not feature requests"

Step 5: Test the Connection

• Click Test Connection to verify it works
• Eloquens AI will discover available tools and display them
• Review the tools to understand what capabilities are available

Step 6: Save Your Connection

• Click Create or Save
• Your MCP connection is now available for use in workflows

Part 2: Associate MCP Connection with Workflow

Once you've created an MCP connection, you need to enable it for specific workflows.

Step 1: Navigate to Your Workflow

• Go to Workflows in the main navigation
• Either create a new workflow or select an existing one to edit
• Progress through setup until you reach the Knowledge AI section

Step 2: Access Connections Tab

• In the Knowledge AI section, click on the Connections tab
• Scroll down to the MCP Integrations section

Step 3: Enable MCP Connection

• You'll see a list of all your MCP connections
• Find the connection you want to use
• Click the checkbox or toggle switch next to the connection name to enable it
• Enabled connections will be marked as active

Step 4: View Available Tools (Optional)

• Click View Tools on any connection to see what capabilities it provides
• Review whether tools are classified as KNOWLEDGE or ACTION
• This helps you understand what the AI can do with this connection

Step 5: Configure Multiple Connections (Optional)

• You can enable multiple MCP connections for a single workflow
• AI will intelligently choose which connection to use based on email content
• Each connection operates independently

Step 6: Save Your Workflow

• Click Save changes or Continue to save your workflow configuration
• MCP Integrations are now active for this workflow

How MCP Integrations Work in Practice

When You Receive an Email

Smart Filter Review: Email passes through your Smart Filter first

Knowledge Enhancement: If AI needs external information to respond:

• AI analyzes the email content
• Determines which MCP connection has relevant information
• Searches connected systems automatically
• Incorporates findings into the response

Action Execution: If Intelligent Routing determines actions are needed:

• AI identifies appropriate actions based on email content
• Selects the correct MCP tools
• Executes actions in external systems
• Includes action results in the response or routing message

What Gets Included

For Knowledge Tools:

• Relevant information retrieved from external systems
• Citations or references to source documents
• Real-time data from connected databases
• Context from related tickets, projects, or records

For Action Tools:

• Confirmation of actions taken
• Links to created tickets or records
• Status updates from external systems
• Results of executed workflows

Managing Your MCP Integrations

To Edit a Connection

• Go to Knowledge Manager → Connections
• Find the MCP connection in the list
• Click the menu icon (⋮) → Edit Connection
• Update settings and click Save

Note: You cannot view or edit saved credentials for security reasons. To change credentials, you must re-enter them.

To Test a Connection

• In the connections list, click Test on any MCP connection
• Eloquens AI will verify connectivity and discover available tools
• Review the test results to ensure proper functionality

To Disable a Connection in a Workflow

• Go to Workflows → Select your workflow → Knowledge AI → Connections tab
• Find the MCP Integration in the list
• Toggle or uncheck the connection to disable it
• Click Save changes

To Delete a Connection

• Go to Knowledge Manager → Connections
• Find the MCP connection
• Click Delete or the delete icon
• Confirm deletion
• The connection is removed from all workflows

Warning: Deleting a connection will disable it in all workflows that use it.

Understanding Tool Classifications

When you view tools for an MCP connection, you'll see classifications:

KNOWLEDGE Tools:

• Marked with a knowledge/search icon
• Used for retrieving information
• Activated when AI needs context to respond
• Examples: search_confluence, get_jira_issue, query_database

ACTION Tools:

• Marked with an action/execute icon
• Used for performing tasks
• Activated when Intelligent Routing determines actions needed
• Examples: create_jira_ticket, update_record, send_notification

AI automatically determines when to use each type based on your workflow configuration and email content.

Best Practices

Configuring Connections

Descriptive Names: Use clear, specific names that indicate what the connection does

Detailed AI Instructions: Provide clear guidance on when and how to use the connection

Test Regularly: Periodically test connections to ensure they remain functional

Secure Credentials: Store connection credentials securely and rotate them periodically

Using MCP in Workflows

Start with Knowledge: Begin by enabling knowledge tools to enhance responses

Add Actions Gradually: Enable action tools once knowledge integration is working well

Monitor Performance: Review workflow results to see how MCP tools are being used

Refine Instructions: Update AI instructions based on how tools are utilized

Multiple Connections

Complementary Systems: Connect multiple systems that serve different purposes

Avoid Overlap: Don't connect multiple sources with identical information

Clear Purposes: Ensure each connection has a distinct role in your workflow

Load Distribution: Multiple connections allow AI to choose the most appropriate source

Security and Access Control

Your MCP integrations are secured through:

Encrypted Credentials: All authentication credentials are encrypted in storage

Workspace Scoping: Connections are private to your workspace

Secure Communication: All data transfer uses encrypted connections

Access Control: Only workflows in the same workspace can use connections

Audit Trail: All MCP tool usage is logged for review

What You'll See in Action

When MCP Integrations are active:

Enhanced Responses: AI responses include information from connected systems

Automatic Actions: Tasks are completed in external systems without manual intervention

Cited Sources: Responses reference where information came from

Action Confirmations: Clear indication when actions are taken in external systems

Workflow Transparency: Full visibility into which tools were used for each email

Common MCP Integration Examples

Atlassian Integration (Jira/Confluence)

Knowledge: Search Confluence pages, retrieve Jira ticket details, look up project information

Actions: Create Jira tickets, update ticket status, add comments to issues

Database Integration

Knowledge: Query customer records, look up order history, retrieve account details

Actions: Update records, log interactions, create new entries

Custom Application Integration

Knowledge: Retrieve data from proprietary systems, search internal databases

Actions: Trigger custom workflows, update system state, send notifications

This feature integrates seamlessly with your Smart Filter, Intelligent Routing, and Knowledge AI capabilities to create powerful, automated workflows that connect your email communications with your broader business systems.