Since its founding in 2014, India-based Postman has made a name for itself as one of the most popular platforms for building and using APIs, with 500,000 organizations now using the service. Like many other SaaS services, the value is now down from $5.6 billion in 2021. But the company is now jumping into the AI space with the launch of an AI agent maker that integrates a large language model. and the Postman API platform, as well as a visual editor that brings the two together to help even non-developers build and test AI agents. At first, jumping into an AI agent might seem like a strange choice for a company like Postman, but as Postman CEO and co-founder Abhinav Asthana noted when I spoke with him before the agent’s launch, to be useful, the agent needs to interact with multiple services. – and the way to do it is through the API. “Our customers started asking about it. ‘Okay, this agent is coming. What are you going to do about it?’ And we see the same pattern integrating this. Whether they use a homegrown system or use some third-party stuff, the API is an important part, because there are two types of API: one is LLM itself, like OpenAI’s API, or maybe the homegrown API. And the second bit is the tool that the agents are going to use. So we’re like, okay, Postman helps developers do all the API and if this comes up as a pattern, we have to do something here,” Asthana explained. Postman agent workflow.Image Credit:Postman He also notes that many Postman users are not developers, who want to use the service to build simple, API-centric applications. “So when AI technology started to emerge, we had to do a lot of testing. We saw things like, ‘okay, does this work?’ And it looks like no code, plus AI, plus agents is a great combo to put together,” Asthana said. The Postman API Network already has an API hub with contributions from more than 18,000 companies, which now also allows users to create AI agents that can interact with the API. To ensure that the agent works as intended, Postman adds testing and evaluation tools to the service to test prompts and inputs for supported models (think OpenAI, Google Gemini, Anthropic’s Claude, Cohere and Meta’s Llama, for example). It then allows users to use Postman Flows, the company’s visual development environment, to build these agents and allow them to interact with the API. Postman LLM evaluation. Image Credit:Postman Currently, Postman’s AI agents mostly focus on enabling back-end operations. While developers can use the service to power front-end services, Postman itself does not yet offer a user interface to directly interact with the agent. Over time, this may change. “We’re all very bullish on agent software, but I think the modality in which we interact with the system will also evolve,” Asthana said. Postman, of course, is not the only company in this space, but above it already has an API hub, Asthana also believes that it can differentiate its services from others – and especially large cloud providers – by offering a better developer experience. “In my opinion, cloud providers always prioritize the cloud consumption of the developer experience, and we are willing to partner with those who actually provide the right developer experience,” he said. The company is still figuring out how best to charge for this new service. The price, Asthana said, will change over time as the team determines how best to determine the value the agent brings.
