Apollo GraphQL today added additional capabilities to the Model Context Protocol (MCP) Server it is making available, to make it simpler for artificial intelligence (AI) applications and agents to declaratively invoke application programming interfaces (APIs).
Additionally, the company has added batch support to Apollo Connectors to its GraphOS platform to reduce the number of REST API calls that might otherwise be required.
Finally, Apollo GraphQL is also adding a self-service subscription option that is intended to encourage developers to experiment with building applications using a platform that streamlines the orchestration of API calls.
Apollo GraphQL CTO Matt DeBergalis said the extensions to the MCP Server that the company already provides will make it simpler for AI agents to retrieve the data needed to perform complex tasks using a platform that is designed to orchestrate both GraphQL and REST APIs.
Enhancements to the MCP Server that facilitate those workflows include a streamable instance of HTTP that scales to levels required by AI agents, additional tool definition capabilities for logically organizing APIs, known as Collections, and an option to now deploy the MCP Server using containers.
The rise of AI applications and associated agents is about to further exacerbate API management issues that many organizations are already struggling to overcome, noted DeBergalis. A survey conducted by Enterprise Strategy Group (ESG) on behalf of Apollo GraphQL, published today finds 37% already struggle with API sprawl, while 40% encounter API security challenges. More than three-quarters (76%) are unable to execute quick changes due to infrastructure rigidity, according to the survey.
While MCP Servers and the agent-to-agent (A2A) protocol developed by Google provide a lightweight framework for access data and enabling AI agents to interoperate with one another, there will also be a need for a platform that streamlines the management of the API calls those agents will be making at scale, said DeBargalis.
Additionally, platforms such as GraphOS will be able to narrow the range of data exposed to AI agents to both improve results by helping to reduce the number of hallucinations that might otherwise be generated and add governance capabilities, he added.
Hopefully, as integrations between platforms based on GraphQL and various frameworks used for integrating AI agents become tighter, the overall robustness of AI applications will steadily increase. Many of the use cases involving MCP Servers and the A2A protocol are still relatively immature simply because of issues involving the amount of data that can be accessed at the levels of scale required, noted DeBargalis.
In the meantime, interest in using AI agents to drive a wider range of new applications only continues to increase. In fact, a Futurum Group report projects that AI agents over the next three years will drive $6 trillion worth of economic value by 2028. Realizing the potential, however, will require finding a way to orchestrate the management of the APIs that AI agents depend on at levels of scale that today are rarely seen in most enterprise IT environments.
