Postman Unveils New AI-Enhanced API Tools
LOS ANGELES — Postman introduced a number of AI-enabled tools and features Wednesday that are designed to simplify API creation and make it easier for developers to build AI agents to automate workflows.
Among the new tools is Agent Mode, an AI-powered assistant that turns requests into actions. It can understand natural language commands and execute tasks such as designing, testing, documenting and monitoring APIs. This can reduce manual tasks down from weeks to hours, Postman said.
Postman CEO and co-founder Abhinav Asthana spoke with The New Stack at POST/CON, the company’s annual user conference, held this week in Los Angeles.
“The key bit here is that essentially every single functionality, including insights, monitors, testing, documentation — every single thing — is exposed to the agent as a tool,” Asthana said. “The agent, prompted by the human developer, can orchestrate things in ways that just could not be done before.”
Postman Introduces an AI Agent Builder
Postman also introduced an AI Agent Builder that integrates with the Model Context Protocol (MCP), which can connect large language models (LLMs) to external documents and tooling.
With AI Agent Builder, developers can create their own multistep agents to automate common API tasks. Agent Builder is best thought of as a suite of solutions that provide three functions, explained Rodric Rabbah, Postman’s head of product.
First, it gives developers the ability to test and evaluate LLMs against prompts so that developers can test LLMs to determine which best fulfills their use case.
Second, it offers an AI tool builder that can turn an API into an MCP specification, which is the standard way for attaching an API to an LLM without having to rebuild, Rabbah said.
Third, it includes a visual drag-and-drop development tool called Flow, which lets developers build applications using API calls to AI. Developers can create the flows and run them in a serverless manner in Postman’s cloud, he added.
Among the internal use cases for Flows that Postman’s team has developed are an automatic workflow to manage laptop procurements and refreshes in Slack, generating security reports for the CISO and a customer support flow that can generate a response to new questions based on historical customer data. Postman offers a template catalog for flows and some AI agents that the company has created in-house.
Improved Observability for APIs
The company also released Postman Insights, a real-time observability tool for APIs. It can track usage across endpoints and versions, detect failure patterns and proactively resolve those issues before they affect users.
Postman Insights supports what the company calls a Repro mode, which reproduces API failures instantly using real-world headers, payloads and authentication tokens. It’s useful for debugging agent interactions in production. In a demo, the company showed how it can fix bugs automatically with a button click.
Asthana explained it’s like drop-in observability for API endpoints that works without any configuration, allowing developers to “drop in a utility in their infrastructure, and using eBPF [Extended Berkeley Packet Filter], it automatically infers these API endpoints.”
The Linux kernel technology eBPF allows users to execute custom programs within the kernel without directly modifying the kernel source code or loading.
There’s a bit of AI/machine learning (ML) magic to convert the traffic into useful API endpoints, he added.
“The advantage in this model is that, unlike traditional observability, where you have to instrument service by service, it can actually infer the ones that you have not instrumented,” Asthana explained. “In that model, it can detect unknown endpoints. It can just give you that picture that you never had before.”
It can also be connected to the development workflow, he added. Most of the time, developers will catch something in their observability system and then have to trace it, log it and then copy and paste those endpoints back into Postman. This eliminates that step.
“With what we call Repro mode, it automatically recreates that request packet and plugs it back into Postman so you can start debugging, testing and just complete your workflow without you doing anything other than clicking a couple of buttons,” Asthana said.
Postman also unveiled Postman Monitors, which complements the Postman Insight offering by allowing developers to set up pre-production tests on the production environment and then run these tests outside the API endpoints. Monitors are synthetic, which means they can proactively test and monitor the performance and availability of applications and APIs by simulating user interactions.
A big part of the Postman news was its integration and use of MCP, Anthropic’s protocol for sharing external information with LLMs. It has integrated MCP with its Agentic AI Builder to turn APIs into callable agent tools. It can also generate MCP servers directly from API collections and test agent behavior with a new MCP request type, the company stated.
As part of its overall MCP strategy, Postman introduced an MCP server network, a curated network for MCP servers that’s hosted on the Postman API platform. Verified publishers can self-publish tools for agents, which makes it easier for developers to discover reliable servers.
Postman has more than 18,000 publishers and 100,000+ APIs on its API Network.
The company is also supporting more enterprise-focused workflow integrations:
- GitHub for real-time collection sync and branch-based governance;
- Jira for context-aware issue creation and tracking from within Postman’s platform without switching to Jira; and
- Slack/Microsoft Teams for improved notifications for teamwide visibility.
Editor’s note: Postman paid travel and accommodations for Loraine Lawson to attend POST/CON.
