App Connections & MCP
Connect your workspace to external services, databases, and APIs
This guide explains how to connect your AdeloopAgent workspace to the outside world, allowing your workflows and agents to interact with external services, databases, and APIs.
Here is a visual breakdown of how standard app connections compare to the Model Context Protocol (MCP):
🔌 App Connections
App Connections are secure links between your workspace and standard external applications.
How App Connections Work
- Secure Authentication: App Connections store your authentication credentials, such as API keys or OAuth tokens, securely within the platform.
- Workflow Integration: Once connected, you can use the App's specific actions and triggers directly in your workflows without needing to log in every time.
- Agent Integration: Your AI agents can be granted access to these connections, allowing them to autonomously interact with the connected services on your behalf.
🌐 Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an advanced feature that takes app connections to the next level. Instead of relying on hardcoded integrations, MCP allows you to dynamically connect to remote servers and instantly expose their custom tools to your workspace.
How MCP Works
When you set up an MCP Connection (typically via an API key), you gain access to a powerful "MCP Server" app in your toolkit. This app provides two primary ways to interact with the remote server:
1. Ask MCP Server (AI-Driven)
This is the most powerful and flexible way to use MCP.
- How it works: You provide a natural language instruction to the AI (e.g., "Query the latest campaign data, calculate the most important KPIs, and explain what changed since last week").
- The Magic: The AI automatically inspects the connected MCP server, discovers what tools are available, chooses the right ones for the job, executes them, and summarizes the results for you in a clear, readable format (like a markdown table).
2. Call Specific Tool (Manual)
If you know exactly what you want to do and don't need AI decision-making, you can use this action.
- How it works: You explicitly select a specific tool exposed by the MCP server and provide the exact JSON arguments it requires.
- Best for: Predictable, automated workflow steps where you want precise control over the data being sent to the remote server.
MCP and Agents
MCP is fully available to your AI Agents! By equipping your Agent with an MCP connection, it gains the ability to dynamically discover and use external tools in real-time to solve complex problems, essentially giving your AI an expandable, unlimited toolkit.