AppDev leaders eye Postman AI tools for API development
Postman's agentic automation could add efficiency and ease of use, plus specialized context and governance guardrails, as AI apps reshape API development
Postman added an "Agent Mode" and other AI features to its software this week, a move welcomed by one of its customers and a large business partner as AI apps alter the evolution and management of APIs.
Similar to general-purpose AI coding agents, Postman's Agent Mode can summarize APIs, diagnose issues and apply updates. However, Postman has trained large language models (LLMs) and developed agents that specialize in API development, said Balaji Raghavan, head of engineering at Postman.
"APIs can get really complex and really big, when you think about, for example, a banking system," he said. "If you just dump everything as context into a model to operate with, you're soon going to have a lot of hallucinations and lose the ability to contextualize down to the specific request that you're trying to operate on or modify. This is where we started building agents that can really focus and solve for specific test, design and development journeys."
Balaji Raghavan
Postman's API Catalog was also among multiple updates to the API testing and development vendor's software this week. It offers a consolidated view of which APIs are in use in development, test and production environments and who owns them. The feature had already been in development at Postman before the AI coding craze hit but moved up the priority list in response to the trend, according to Raghavan.
"With agentic AI, there are going to be more APIs needed because you don't want broad exposure of your APIs -- there will be more APIs generated for agents to operate [safely] with back-end systems," he said. "Internally, we were seeing an explosion of the number of APIs through either direct HTTP web sockets or even MCP [Model Context Protocol], and we quickly realized that there needed to be some way for leaders to start governing that."
Postman also built out further AI-focused integrations into its platform with this release. They include:
Git workflows that manage API specs, collections, tests, mocks and environments directly in developers' git repos and local file systems;
MCP servers from Atlassian, Amazon CloudWatch, GitHub, Linear, Sentry and Webflow;
An integrated API distribution and publishing mechanism for documentation, workflows, sandboxes and SDKs in one place.
Context engineering is the real differentiator for agentic AI platforms and that's where Postman makes its bet.
Torsten VolkAnalyst, Omdia
The major challenge for Postman will be rising above the noise in an overstuffed marketplace for AI agent tools, according to Torsten Volk, an analyst at Omdia, a division of Informa TechTarget.
"Context engineering is the real differentiator for agentic AI platforms and that's where Postman makes its bet. … [It] already owns the API catalog, the specs, the test results and increasingly the production behavior data that agents need to make informed decisions," Volk said. "The open question is whether Postman can hold on to that context advantage as IDE-native AI assistants, observability platforms and platform engineering toolchains all race to assemble the same picture from [another] direction."
Specialized API AI piques customer's interest
One Postman customer said he sees the appeal of specialized agents for API development, where he was skeptical of last year's market buzz around vibe coding. This focus, along with the governance features built into Postman, makes its AI agents more appealing to enterprise teams, according to Ben Heil, a principal engineer at an HR software company based in the Midwest that he requested not be named.
"API quality and consistency are more important than ever because of AI -- this is in contrast to the 'vibe coding' movement that we've seen over the last couple of years," Heil said. "Without quality and consistency standards, the API sprawl problem created by AI acceleration will create mountains of technical debt that will smother innovation and time to market for teams."
Heil said he plans to use Agent Mode at his company and said he sees it as an improvement over Postman's previous LLM assistant called Postbot.
"The problem with chatbots is that they are single-turn, which limits their ability to do more than generate individual artifacts," Heil said. "The single-turn nature also limits its ability to make cohesive design 'decisions' because the context between requests needs to be reestablished each time. With a multi-turn agentic system, the possibilities expand."
Heil also said the new git integration is a smart move for Postman, making it "a more ergonomic fit with the way most software teams operate today, especially with tools like GitHub Copilot or Claude Code, which are also operating directly off of git repository context and instructions."
PayPal adds Agent Mode to Postman collection
Postman business partner PayPal operates a publicly available collection of APIs in Postman's hosted collections system, which third-party developers download and fork to add workflows -- such as checkout processes or payment buttons -- to their own apps and websites. Some 100,000 developers have forked PayPal's collection to date, making it the most popular on Postman's platform, according to Bo English-Wiczling, vice president, global developer relations, at the digital payments processor.
Bo English-Wiczling
"Agent Mode is already enabled [for that collection] … in integration with our MCP server, so you can just say, 'Hey PayPal, create an invoice for me for $100 and send it to this individual,' and it will actually generate that invoice for you, and we'll send a note to that email address that you indicated," English-Wiczling said.
This process of natural-language programming has the potential to make APIs much easier to use and shift how online shopping is done, she said, by agents rather than human developers. But behind the scenes, it also takes a careful approach to API development that makes APIs clearly readable by AI agents as well as humans.
"You have to give [agents] these very specific step-by-step instructions so that they can integrate properly," English-Wiczling said. "For example, in a lot of text files, there's a markdown file for a lot of sites like 'llms.txt', that gives the agent specific instructions on how to read the thing."
PayPal uses a different product, APIMatic, internally to develop APIs, but partners with Postman to make its APIs widely accessible through its collections. The company also employs AI agents and LLMs from various sources to develop, test and document APIs. Now, a more unified approach to AI for API development across these different tools is beginning to take shape at PayPal, English-Wiczling said.
"We're working on making things like our APIs and our SDK [follow] a consistent mental model [with] different technologies that developers like to use," she said. "We're launching into multiple different areas that developers expect us to be at, but maybe they don't see us at, like frameworks, for example, and IDE tools."
Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism. Have a tip? Email her.
Balaji Raghavan
Bo English-Wiczling
