Postman today revealed it has developed artificial intelligence (AI) agents for its platform for designing and building application programming interfaces (APIs).

Announced at the POST/CON 2025 event, the Agent Mode enables developers to, via a natural language interface, automate entire multi-step workflows spanning the designing, testing, documenting and monitoring of a single API or a logical group of APIs, known as collections.

Postman CEO Abhinav Asthana said unlike an AI copilot, Agent Mode natively executes tasks in a way that understands the actual intent of an API developer.

The overall goal is to reduce the amount of time required to build and deploy an API from weeks to a few hours using a set of repeatable workflows created by an AI agent, he added.

Additionally, developers will be able to integrate the Postman agent with other AI agents via a model context protocol (MCP) server that Postman added last month.

Finally, Postman is also committing to make it possible for developers to alternatively employ AI agents they create to build, develop and test APIs. Postman Insights, now available in a more open beta, will then provide the tools required to track usage of any API created in real time, noted Asthana.

It’s still early days so far as usage of AI tools to build APIs is concerned, but reliance on them within AI applications is already pervasive. Based on data collected from the Postman API Platform between May 2023 and April 2025, developers made more than 7 million API requests involving AI. The OpenAI platform accounted for 60% of that activity, but API traffic involving Meta Llama and Google Gemini grew by a factor of seven and five, respectively.

In total, there are more than 40 million developers that make use of Postman tools to build APIs for 500,000 organizations. The Postman API Network is being used to expose more than 100,000 of those APIs to drive multiple application services. As AI agents make it simpler for both developers and now non-technical end users to build and deploy APIs, the volume of activity is only going to continue to exponentially increase, noted Asthana.

Each organization will need to determine how extensively to rely on AI agents to build and deploy APIs, but it will, if necessary, be possible to rollback any tasks completed by an AI agent, said Asthana.

The one thing that is certain is that API documentation should substantially improve. Most application developers don’t especially enjoy creating documentation, which can be a significant issue whenever a DevOps team needs to troubleshoot an application.

Hopefully, an AI agent trained to develop APIs will lead to higher quality. APIs, in particular, have become the focal point for cybersecurity attacks because in addition to enabling the exfiltration of data they can also be a vehicle for corrupting business logic.

In the meantime, however, IT leaders in the months ahead might want to revisit not just how APIs are created, but also managed, secured and updated in the age of AI.