Setup MCP Action

This guide walks you through connecting an MCP server to Enjo and creating an AI Action from it, so your AI Agent can call external tools like Notion or Stripe during a live conversation.

Note: This guide covers two steps: connecting the MCP app in Settings, then creating an AI Action from it in your AI Agent. Both steps are required.

Model Context Protocol (MCP) is an open standard that exposes tools and contextual data to large language models through a uniform, server‑based interface. By supporting MCP servers, Enjo AI Agent Studio lets your agent call external tools such as Notion or Stripe without bespoke integrations.

By acting as an MCP Client, Enjo AI Agents can:

• Discover tools and APIs exposed by an MCP server.

• Execute those tools directly as AI Action or as part of AI flows.

Prerequisites

1. Role permissions: Owner or Developer in the target workspace.

2. Enjo dashboard access for your environment (.enjo.ai/dashboard).

3. MCP server credentials:

   1. For pre‑built servers: OAuth or API token for the target SaaS (e.g., Notion, Stripe).

   2. For custom servers: Server URL and optional Auth Token.

4. Allowed outbound traffic from Enjo Servers to the MCP server endpoint

Part 1: Connect the MCP app

Step 1: Open Apps in Settings

1. Log in to the Enjo dashboard.

2. Navigate to Settings and select Apps.

3. Click Connect Apps in the top right corner.
   
   

Step 2: Select and connect your MCP server

Choose the type of MCP server you're connecting:

Pre-built MCP app (e.g. Notion, Stripe)

1. Find your app in the list. Pre-built MCP servers are labelled as MCP.

2. Click Connect.

3. Complete the OAuth or token-based authentication flow.

Custom MCP server

1. Select Custom MCP.

2. Enter a name for the connection, the MCP server URL, and an optional Auth Token.

3. Click Connect.

💡 Tip: If your MCP server requires IP whitelisting, contact your vendor to request that Enjo's server IPs are added before attempting to connect.

Once connected, the MCP app is available to use across any AI Agent in your workspace.

Part 2: Create an MCP AI Action

Step 3: Open the AI Actions tab

1. Navigate to AI Agents and select the agent you want to configure.

2. Click the AI Actions tab.

3. Click + Add AI Action and select Create from MCP Server.
   

Step 4: Select your MCP app and choose a tool

1. Select the MCP app you connected in Part 1. Enjo fetches the list of available tools automatically.

💡 Tip: If tools don't load, click Fetch Tools to retry. If the issue persists, confirm the MCP server is reachable and your credentials are valid.

1. Choose the tool you want to expose as an AI Action, for example Create GitHub Repo or Move Notion Page.

2. Click Next.

Step 5: Configure and save the action

1. Review the auto-populated AI Action Name and Output Description. Update the description only if the default does not accurately describe what the action does.

2. Click Test to validate the action against the connected MCP server.

3. Click Save.

What happens next

The MCP AI Action is now available in your AI Agent. The agent will invoke it when a user's request matches the action's description. You can view and manage it under the AI Actions tab in your agent.

Troubleshooting

Can't connect to the MCP server. Some MCP servers require Enjo's server IPs to be whitelisted. Contact your MCP server vendor to request this if needed.

Tools don't load after connecting. Click Fetch Tools to retry. If tools still don't appear, confirm the server is online and your credentials haven't expired.

The MCP server goes down after setup. Tool calls will fail and tool discovery will not be possible. Any AI Actions or Flows using MCP tools will not work until the server is back online.

The agent isn't calling the action. Review the action's description and add question variations that reflect how users phrase this type of request. See [AI Actions Troubleshooting] for more guidance.