玩偶姐姐

Keith Shaw: Hi everybody, welcome to DEMO, the show where companies come in and show us their latest products and platforms. Today, I'm joined by Taylor Pechacek. He is the Head of Product for API Collaboration Core at Postman. Welcome to the show, Taylor.

Taylor Pechacek: Hey, great to be here. Keith Shaw: So tell us a little bit about Postman, for those who might not be familiar, and then what you're going to be showing us today. Taylor Pechacek: Awesome. Yeah, so Postman is the world's leading API collaboration platform.

It helps developers and software teams around the world manage the entire API lifecycle. That means planning and designing APIs, testing them, building and deploying them, and managing their ongoing maintenance. We really cover the full end-to-end lifecycle.

It’s widely used—about 500,000 organizations already use Postman, with 35 million developers and 98% of the Fortune 500 relying on it.

Keith Shaw: Taken off, right? Taylor Pechacek: They have, yeah—definitely. Keith Shaw: Today, you're going to show us some agentic AI capabilities. Taylor Pechacek: Yes, very excited about this. We're constantly improving the product, and today we have a major release around how to build agents on the Postman platform.

Keith Shaw: That’s a huge topic. So who is this really designed for? I'm guessing software developers, but are there others within a company who could benefit from this? Taylor Pechacek: Great question. It's really aimed at our core market—developers.

But there are many types: backend, frontend, QA, platform engineering, DevOps. And since we're a collaboration platform, APIs are central to modern software development, which also includes product managers, designers, solution engineers, and sales teams. Agents will also help non-developers start building by making it more accessible.

It expands beyond our developer core to help a broader audience improve their workflows.

Keith Shaw: So what are some of the problems you're solving? I'm assuming it's: "How do I build an agentic AI?" Instead of using a different platform, you're already working with Postman. Is that one of the key benefits? Taylor Pechacek: Exactly.

One big benefit is that Postman provides a suite to discover the right APIs and LLMs to use in your agents. We’ll talk about that. You can test functionality, integrate, and build through our Flows experience—all in one platform.

Otherwise, you’d juggle dozens of apps, manage multiple keys, and deal with security issues. We bring it all into one place. And yes, we're leveraging existing Postman features while introducing new ones to support agent-building.

Keith Shaw: And I imagine IT leadership hears “another platform” and cringes. But since they likely already use Postman, this just adds functionality. Taylor Pechacek: Exactly. This leverages internal APIs and also connects to hundreds of thousands of public APIs.

Agents can access tools like Slack, Notion, UPS, and more—combining LLM reasoning with traditional systems to act on workflows.

Keith Shaw: So let’s dive into the demo. Taylor Pechacek: Great. What I’ll show you involves the agent framework: building agents, discovering APIs and models, evaluating and testing them, and keeping them running reliably.

Let’s jump into Postman. First, I’ll orient you around the core workspace. We have a made-up company called ShelfWise. I’m a big reader—love Goodreads. ShelfWise helps people track books and get recommendations. We store all APIs this company uses in collections.

Postman is known for allowing GET and POST requests through the interface. For example, you can fetch Slack channel history directly via Postman.

Using Postman’s UI, I can hit any internal or external API. We support multiple protocols like HTTP, GraphQL, and gRPC, which are popular in microservices. With the rise of AI, we’ve introduced a new request type: LLMs. Select “AI request,” and we offer options like OpenAI, Google, Anthropic—more to come.

Being able to swap out models easily is crucial. Let’s choose OpenAI. I’ll enter my API key and start playing. It’s already generating responses about what Postman does—probably better than I explained it!

You can view token usage, costs, response times—all important for evaluating model efficiency and integrating LLMs into software.

Now, say you want to evaluate multiple models across providers. For example, we’ll test five OpenAI models using a collection runner. You can run these tests manually or integrate into your CI/CD pipeline.

In this example, we’ll prompt each model with “What does Postman do?” and ensure the token count stays below 2,000.

Postman hits all APIs, runs tests, and shows which model performs best based on cost, speed, and complexity.

Now let’s talk visualization. You can compare models side by side—response times, quality, cost—helping teams make smarter decisions.

Next, let’s build an agent using Flows. This is a visual designer for creating workflows with embedded LLM reasoning. Here’s our incident management example. We use Slack and Notion.

The agent opens a Slack incident channel, gathers messages, evaluates them, and decides whether to create/update a Notion page and alert leadership—all autonomously.

Here’s the magic: within a module, you instruct the LLM in natural language—e.g., “Given these Slack messages and Notion properties, decide whether leadership should be notified.”

You don’t know what it will decide—it depends on the input. Fine-tuning matters. Let’s test this with a real scenario.

We have an incident affecting 100% of users—customers can’t submit payments for 3 hours. Sounds serious, right?

We paste that message manually, but it would normally be automated. The agent fetches the message, thinks, evaluates, and creates a Notion page. It also notifies engineering with a crafted message. All of this links to the database summary.

It determined the severity itself based on the content. We didn’t tell it outright—it inferred from context.

Now let’s say the issue is resolved. I’ll update Slack messages accordingly.

We run the agent again. It fetches updated Slack messages and determines we don’t need to alert leadership. So this isn’t a one-time action—the agent continues monitoring Slack over time.

You can configure it to act on every message or at timed intervals.

Here’s the cool part: Postman already has all these APIs on its network. Let’s say we want Slack conversation history. Normally, you’d hunt through Slack docs. But here, we just search for “channel messages,” and Postman builds the request for you.

So instead of building everything from scratch, you’re leveraging high-quality APIs curated and promoted by companies like Slack and Notion. Keith Shaw: Awesome.

So where can people go for more info? Is this available now? Taylor Pechacek: You can start building today at postman.com. Just search for “Postman AI Agent Builder.” You’ll find collections, examples, and Flows to fork and use right away.

Keith Shaw: Cool stuff. Taylor, thanks for the demo. Taylor Pechacek: Awesome—thank you!

Keith Shaw: That’s all the time we have for today’s episode. Be sure to like the video, subscribe to the channel, and leave a comment below. Join us every week for new episodes of DEMO. I’m Keith Shaw — thanks for watching!