Postman today opened its catalog for application programming interfaces (APIs) in a way that makes it possible to manage specifications, collections, tests, mocks, and environments directly from within Git repositories and local filesystems.

Balaji Raghavan, head of engineering for Postman, said that capability provides the added benefit of making it simpler to expose the code used to create an API to an artificial intelligence (AI) agent.

At a time when more software engineering teams are relying on AI tools to write code, making it simpler for AI agents to directly see how the code is used to construct an API without having to invoke an additional tool streamlines workflows, said Raghavan. Instead of having to invoke a separate AI assistant, the intelligence layer that Postman is exposing provides full visibility into specifications, tests, environments, and real production behavior within the context of a Git-based workflow, he added.

According to a recent State of the API Report from Postman, 89% of developers are already using AI. Postman has already developed its own AI agents for designing and building application programming interfaces (APIs), but is now moving to expose its API Catalog to third-party AI agents at a deeper level.

Regardless of approach, the overall goal is to make it simpler for AI agents to select which API to invoke based on the functions exposed in the underlying code while at the same time dramatically reducing the amount of time required to build and deploy an API, said Raghavan.

AI agents, in general, depend heavily on APIs to discover what capabilities exist within an IT environment. The better those APIs are in terms of the quality of the code created and the documentation provided, the more likely it becomes that the code being generated by an AI agent will behave as expected. As such, many organizations will be revisiting how their APIs have been built and deployed in the age of AI, noted Raghavan.

In total, there are more than 40 million developers from 500,000 organizations that are using the Postman platform to create APIs. The depth at which those APIs are managed and secured has tended to vary widely from one organization to another. More than a few organizations have rogue APIs that were documented by developers or zombie APIs that are no longer being maintained but are still accessible. The challenge is that AI agents will discover and invoke any API they find in an IT environment unless specifically trained to only access APIs that are stored in some type of central repository.

It’s not clear how rapidly organizations are adopting AI agents to specifically build APIs, but it’s also only a matter of time before AI agents are routinely called upon to deploy, manage and secure them. In fact, the overall state of API management should improve in the age of API because many application development teams still lack any formal processes for managing them.

In the meantime, however, those same organizations should expect to be building and deploying more APIs than ever in an era where AI agents are going to show a strong preference for invoking services that enable them to easily understand the underlying code.