API Simulation Reduces MCP Server, Microservices Overload
APIs have become so ubiquitous that it’s easy to underestimate the work they can create. It turns out, mo’ APIs, mo problems, to borrow from The Notorious B.I.G.
It’s a problem developers need to consider as they leverage MCP servers, advised Uri Maoz. As the co-founder and CEO of Wiremock, an open source HTTP mocking tool and cloud platform, Maoz has observed the problem of proliferating APIs already in enterprise customers that adopted microservices, he said.
“A lot of enterprise customers say, ‘We moved to [a] microservices environment, and now we realize we have a problem. We cannot develop and test because it’s so limiting,’” he said.
A similar scenario is likely to unfold as more organizations develop and adopt Model Context Protocol (MCP) servers, Maoz said.
“They don’t anticipate the problems that it will cause them later on,” he said. “People will start at some point paying attention to these challenges, and then they will need the ability to run in an isolated way.”
API Proliferation in Enterprises
Increasingly, WireMock sees companies struggling to manage the growing number of APIs in the modern enterprise, he said.
“You have an organization move to a microservices environment and move to consume a lot of third-party APIs, cloud services, and now AI, and the result is that the environment of the developer and the tester now is much more complex, because they have a lot of external and internal dependencies that they don’t have control of,” Maoz said. “Then when you try to develop against these dependencies and testing, it’s becoming really challenging. … And it’s not stable.”
MCP servers will add to API proliferation, he contended. In part, that’s because everybody is pushing MCP servers now. It’s also because not only are you making an API call to an MCP server, the MCP server itself makes API calls that need to be tested against.
“Let’s say you’re using an MCP server that is calling to several third-party APIs. You want to test the workflow, “ he said. “How do you even test it? The same problem that we see today in the market will still be there and with higher volume with the growth of MCP.”
Mocking as a Solution
API mocking (which simulates specific API interactions) or API simulation (which goes a step further, in mimicking an API to provide for more complex testing scenarios) solve the proliferation problem. Developers can mimic the API dependencies, both internal and external, while developing and testing the product, he explained.
“The developer or the tester can control how the request and response will look and can test different scenarios and different workflows,” he said.
Mock servers simulate APIs for development and testing so that you aren’t calling the actual API while working on your application.
WireMock offers an open source model and a for-pay cloud platform to support scale and enterprise. Their API simulation tool creates the mocked API either with its user interface for defining how the API should function, or by recording real calls and traffic to the actual APIs. In the last scenario, it behaves as a proxy and listens to the traffic and calls created by the real API. Once you finish the recording, WireMock can create all the simulations that represent the traffic, Maoz explained.
“Then basically, the workflow will be that the developer or tester, instead of working against the real API, they will mock again the simulated API,” he said.
API simulation differs slightly from mocking, which requires you to create the API mock at the code level. This means every change made to the mock API requires it to be rebuilt.
“The developer or the tester can control how the request and response will look and can test different different scenarios and different workflows.”
— Uri Maoz, co-founder and CEO of Wiremock
“There are mocking tools that enable you to do the mock, but then it’s static,” he said. “It means that every time you call a mock, you will get the same value.”
With API simulation, the API behavior saves to the cloud and the only change the developer needs to make is to the base URL, he said.
“Wiremock Cloud enables really comprehensive, dynamic responses,” he said, explaining that it uses stubs to do so. A stub is like the use case or an API endpoint.
“You can create one stub that is dynamic and will know to return dynamic and different responses based on the stuff that you will send in the request,” he said. “This is very important, because now you can create a lot of different use cases with much less maintenance.”
Use Cases for API Simulation
For frontend developers, this means developers don’t have to wait for the backend to create an API before they can begin to test how an app works with specific functions an API might provide.
But there are other uses as well, such as when a developer wants to test a negative situation where there’s a failure or delay of the API. API simulation can mimic that process easily, he said.
“You have full control over how the API will behave,” he said. “Otherwise, how would you do that when you work with the real API or even with the sandbox?”
Another use case is in the financial space, where a developer might want to simulate a banking transaction where a balance is updated because of a deposit or withdrawal.
“This is something that, for example, mock enables and if you don’t have that, then it’s very hard to really simulate the behavior of the real API,” Maoz said.
Finally, the simulation can also be used as a sort of Figma for prototyping a new API, letting developers specify how they want an API to work and generating the OpenAPI specification from that information. Developers can import an OpenAPI specification or use WireMock’s MCP Server integration to create APIs.
“[You] start the prototype process of the API in WireMock Cloud by creating different mocks of how the API should behave,” he explained. “What we do from that point is that we generate for you the open API specification.”
There’s also a validation mechanism that alerts developers if there are discrepancies between the mock and the OpenAPI spec to avoid “mock drift,” which is when the real API changes but the mock doesn’t, he added.
