How to Make Your APIs Agent-Ready With MCP Bridge

Reading Time: 15 minutes

AI agents are only as powerful as the data and systems they can access. As a developer, you’ve likely spent years building out a robust library of REST APIs that handle everything from processing refunds to checking inventory. The goal now is to let AI agents use that work without requiring you to rewrite your entire backend or maintain a dozen standalone servers.

Model Context Protocol (MCP) is the open standard that makes this possible, but manually building and hosting individual MCP servers for every use case creates significant technical debt and operational overhead.

In this guide, let’s take a look at how to use MuleSoft’s MCP Bridge to quickly enable MCP for your existing APIs. We’ll move from REST API endpoints to governed, agent-ready tools in just a few minutes, treating MCP as a clean interfacing layer rather than a new infrastructure project.

Why use a managed MCP strategy?

While you can write a custom MCP server for every API, doing it at the gateway layer with a managed platform provides a few major practical advantages:

• Reuse your existing assets: You don’t need to rebuild logic. Your existing integrations, security policies, OAuth scopes, and rate limits remain the source of truth. You’re just adding a new “agent-friendly” context
• Curated toolsets: Agents don’t need to see your entire technical API surface. MCP Bridge lets you pick the specific, high-value endpoints that make sense for a business process and expose them as clean “tools”
• End-to-end visibility: By managing your MCP traffic through a gateway, you get a clear audit trail. You can see exactly which agents are calling your systems and monitor for any unexpected behavior in real time

What is MCP Bridge?

MCP Bridge is delivered as a pre-configured set of automated policies on MuleSoft Flex Gateway via a guided user experience in API Manager, allowing organizations to enable MCP without changing existing APIs or deployment models. The MCP servers created by MCP Bridge live at the gateway layer. Think of it as a translation engine for the agentic era.

It handles the boilerplate of translating the agent’s MCP requests into the standard REST calls your backend expects. More importantly, it provides a simple interface to add semantic metadata – the natural language descriptions that LLMs need to understand exactly when and how to use your tool.

Tutorial: Making a shipping tracking MCP Server

In this example, we’ll take a Vendor Shipping Tracking API and an ERP System API and transform some of their endpoints into governed MCP tools. This allows an AI agent to answer “Where is the vendor’s shipment?” inquiries autonomously, pulling live data directly from your systems and making updates to the ERP system if there’s a change in shipment status.

Prerequisites: You should have a basic working knowledge of Anypoint Platform before going through this step-by-step guide as you will need to have APIs already deployed in your Anypoint org. 

• Anypoint Platform account with following permissions:
  • API Manager:
    • API Creator
    • Deploy API Proxies
    • Manage APIs Configuration
    • Manage Policies
  • Runtime Manager
• APIs published and managed in Anypoint API Manager in your environment
  • You can use any API at your disposal. If you want to follow along, you can find the mock APIs at this Github repository
• A deployed Managed Flex Gateway and its URL 
• MCP Inspector (or a client like Claude Desktop or Postman) 

You can follow along with the steps outlined below or walk through with this interactive guide. 

Step 1: Create an MCP Server Instance

1. Log into Anypoint Platform (anypoint.mulesoft.com)
2. Navigate to API Manager and go to the Agent and Tool Instances tab.
3. Click Add and select MCP Bridge.

You can also start the process by going to the API Instances tab, choosing the API you want to expose as an MCP server, and under Actions, choose Add to new MCP Instance.

Step 2: Configure runtime settings 

Choose a Managed Flex Gateway you want to deploy the MCP instance to and click Next.

Then, select the API(s) you want included in the MCP instance. You can choose as many as you would like which allows you to curate a specific set of API calls that are useful for a certain process. Only the APIs cataloged in Anypoint Exchange are visible. 

In our example, you will choose horizon-expeditions-netsuite-api and Vendor Shipping Tracking API. Next, create an MCP asset name such as Vendor Shipment MCP and click Next.



Next, configure downstream: This is your public-facing endpoint for agents. Assign a unique Base path such as /vendor-shipping. Add the consumer endpoint under advanced settings. This will make it easier for other developers to get the MCP server’s url to incorporate in their agentic projects. ([flex gateway url]/vendor-shipping/). Click Next after you are done.



Then, configure upstream: This is where you can specify an API’s URL. MCP Bridge will automatically fill out the URL that is specified in the API specification. Configure for both APIs with your url and click Next. 



Step 3: Map API endpoints to MCP Tools

Now, let’s map our API endpoints to MCP tools. An API endpoint is designed for developers; an MCP tool is designed for LLMs. This step is where you give the agent instructions how to use certain API endpoints.

1. Select API Endpoint: Using the UI, select the API, operation method and resource. 

For the first MCP tool, choose:
Select API: Vendor Shipping Tracking API
Method: GET
Resource: /tracking/{trackingNumber}

You’ll notice that you will get multiple available resource options based on the method you choose.

2. AI tool name: MCP Bridge will autopopulate the name based on the API specification. Update the name to use a clear, action-oriented name (e.g., get_shipment_status), if you would like.



3. Tool fefinition: Give the agent specifics on when to use the tool and why. Example prompts from the users will also give the AI agents context on how to use the tool. 

Instead of a technical summary, write it for the LLM: 

"Retrieves real-time shipping status and location history for a specific package using its unique tracking number. Use this tool when a user inquires about the current progress, estimated delivery date, or 'whereabouts' of an order. The tool returns a structured timeline of carrier events, from origin pickup to final delivery."

You can use your favorite LLM to help you generate a description that is helpful for an AI agent. For example, I used the prompt: “Create a description for an MCP tool that is helpful for an agent that has the name get_tracking_trackingnumber” and shared the API specification file to get the above description. 

4. Input schema: This section is autogenerated based on your OAS API spec to let the agent know what information needs to be provided by the user before the agent can use the tool. If you need to make a change to the input schema, you can update the JSON code directly or go to the HTTP mapping tab. 



5. In the MCP tools and mapping section, click Add Tool to add as many tools you would like. Try to only add the essential, unique tools as the more tools the agent has exposure to, the longer it will take them to process which one would be best fit for the task.

For the use case, create a second tool with the following information: 

• Select API: horizon-expeditions-netsuite-api
• Method: POST
• Resource: /shipments
• AI tool name: post_shipments
• Description: “Updates or creates shipment records in the central database. Use this tool to log new outgoing packages, update shipping statuses (e.g., ‘In Transit’ to ‘Delivered’), or modify carrier details for an existing order. This ensures that the system of record remains synchronized with the physical movement of goods.”

Click Next once you’re done configuring the tools.

6. After checking that your configurations are correct, click on Save & Deploy.

Step 4: Test and validate

Once the instance status shows as “Active” copy the consumer endpoint URL.

1. Open the MCP Inspector and set the Transport Type to Streamable HTTP. Paste your URL and click Connect.
2. You should see your tool and its description appear in the tool list. Click on get_tracking_trackingnumber.
3. Enter a sample tracking ID, 776648790327, and click Run. If the Tool result is a success, your agent is now officially empowered to track shipments.



Building for the future

Making your APIs agent-ready isn’t about building more infrastructure; it’s about making your existing work more accessible to the next generation of software. By using MCP Bridge, you can bypass the manual coding of servers and focus on building a curated, secure, and scalable library of tools for your AI agents.

Ready to get started?