CV Deep Dive

Today we’re talking with Abhinav Asthana, co-founder and CEO of Postman — the ubiquitous API platform used by over 40 million developers and 500,000 organizations around the world. Abhinav began coding as a kid in India, built the first version of Postman as a side project to help debug APIs, and has since turned it into one of the most foundational tools in modern software development.

Now, he’s steering Postman through its most radical transformation yet: from an API collaboration platform into a full-stack AI agent development environment. In this conversation, we go deep on his vision for agentic software, the future of developer tools, and why designing APIs — not UIs — may be the most important product decision of the next decade.

Key Takeaways

- Your APIs are your new moat – "An agent that only provides information isn’t actually involved in helping you complete your task." It doesn’t have access to the latest data and tools to function. APIs define what an agent can touch and modify — and Postman already hosts the internal, public, and partner APIs most developers rely on.

- MCP: the next REST? – With Model Context Protocol (MCP), developers can create, test, and publish AI-controllable APIs as easily as REST endpoints. The public MCP Catalog offers prebuilt integrations for Stripe, Notion, Google Maps, and more.

- Agents, not just assistants – Postman’s new AI layer isn’t just about chatbots. Tools like AI Agent Builder and MCP enable developers to build autonomous agents that reason, take action via APIs, and integrate directly into workflows.

- Drag-and-drop, no devops required – Postman’s AI Agent Builder gives developers a visual canvas to chain LLMs and deploy tools into multi-step agents, no infrastructure required. It’s already powering dozens of live workflows inside Postman itself.

- Built by and for developers – Unlike agent frameworks spun out of research labs, Postman’s approach is grounded in a decade of developer feedback and iteration. Its agent ecosystem is natively integrated into the workflows teams already use and are familiar with.

Start building agents today with Postman’s MCP catalog or visit their blog for the latest product updates. Join Discord for discussions on all things AI, APIs, and MCP.

Postman raised a $225M Series D in August 2021 led by Insight Partners, with backing from CRV, Nexus Venture Partners, and Coatue.

Let’s dive in ⚡️

Read time: 8 mins

Our Chat with Abhinav 💬

Good to have you Abhinav! A lot of exciting AI breakthroughs to talk about but first off, give us an idea of your journey and what led you to co-found Postman?

Thanks for having me! I’m Abhinav Asthana. I’m originally from India and I’ve been drawn to technology ever since I was a kid. I started coding back in sixth grade—Visual Basic, Turbo C, the works. That curiosity has stuck with me ever since.

I’ve also always known I wanted to build something of my own. In 2010, right after college, I co-founded a startup that built a collaborative programming environment for education. It was my first real exposure to what it means to ship software and learn from real users. And along the way, one thing kept standing out: the growing complexity of APIs.

At that point, I realized that any real-world application—anything beyond something like a toy calculator project—needed to talk to APIs. But working with APIs was frustrating. There weren’t any good tools out there to test or debug them. That’s when I built the first version of Postman. Initially, it was just a side project—a simple Chrome extension I hacked together to help myself and a few friends at work. But it started spreading quickly, and I began to see just how big the pain point was. That’s what led me to double down, start the company, and build Postman into what it is today: a complete platform for API development and collaboration.

And now, with AI transforming how software gets built, we’re pushing even further—rethinking what developer tools can be in this new era.

Every developer knows Postman, but not necessarily Postman’s new AI features. Could you give us a product breakdown?

Absolutely. So in short: the core platform supports everything API, while our new AI features supercharge 1) your Postman workflow with Postbot and 2) building autonomous agents with AI Agent Builder and the MCP ecosystem.

At Postman, our product suite is organized into two main layers. First is our core API platform, which includes everything our users know and love—API Collections, Workspaces, the API Client, Flows, Mock Servers, Monitors, and our command-line tools and integrations.

Recently, we've introduced a new AI-focused layer on top of our core platform, designed to accelerate productivity and enable developers to easily build AI-driven workflows.

Within this new AI layer, there are two key tracks:

1) Use AI to help users in the Postman app: Postbot, our AI-powered assistant built directly within the Postman app. Postbot is your everyday AI helper to assist in the tasks you used to do manually in the Postman app: write tests, debug API requests, generate documentation, visualize responses, etc. It uses Postman-specific knowledge to assist users efficiently.

2) Help users build AI Agents: These are features for users who want to build agentic workflows without extensive coding.

- AI Agent Builder: The heart of this is the AI Agent Builder—a drag-and-drop canvas and evaluation dashboard that lets you visually chain together large language models (LLMs) and API tools to build multi-step agents. These agents can perform real tasks like calling APIs, interpreting results, comparing models, and taking actions based on context.

- MCP Client: Model Context Protocol (MCP) is now integrated as a first-class request type within Postman, similar to REST or GraphQL. Developers can easily create and test MCP Requests right in the Postman UI, just like they would with REST or GraphQL.

- MCP Server Generation: To build tools that agents can call, we've released the AI Tool Builder. This lets you turn any existing API into an MCP server with a single click. This allows you to expose your APIs in a format directly usable by LLMs.

- MCP Catalog: A public workspace of curated, ready-to-use MCP servers for popular APIs like Stripe, Notion, and Google Maps. Think of it as an npm registry for AI tools. These are verified so developers don’t have to worry about reliability issues.

- Our Postman MCP: Finally, we also offer our own Postman MCP Server, which lets LLM agents programmatically create Postman collections, generate and edit requests, seed example responses, tag workspaces, and more—essentially giving them control over the Postman platform itself

2025 is undoubtedly the year of the Agent. In your December 2024 blog post, you concluded that “the power to deliver agentic AI solutions lies in their APIs”. What do you mean by that?

The core idea is that while large language models (LLMs) can process and generate information, their true potential is unlocked when they can interact with external tools and systems. APIs serve as the standardized interfaces that allow these models to perform actions, such as summarizing Slack messages or executing tasks across different platforms.

However, granting LLMs unrestricted access to all tools can be risky, especially when it comes to actions that might have unintended consequences in production environments. Therefore, it's crucial to control the scope of what these models can access, and APIs provide a structured way to define and manage this interaction surface.

In essence, for AI agents to move beyond passive information processing and become active participants in workflows, they need the ability to perform actions through APIs. This approach not only empowers agents to be more effective but also ensures that their capabilities are governed and safe. As the ecosystem evolves, protocols like the Model Context Protocol (MCP) have emerged to facilitate seamless integration of APIs into agent applications, further solidifying the role of APIs as the backbone of agentic AI solutions.

Walk us through Postman’s AI Agent Builder. Which use-cases do you think are interesting and what should developers experiment with first?

Actually, we're actively using AI Agent Builder at Postman today to deploy agents for our own workflows. Currently, we have about 20 agents running in production across various functions like engineering, finance, security, sales, and product management. Common examples include summarizing user feedback into actionable task lists, security workflows, and sales support automation. Developers can check out our prebuilt examples to get started easily.

We started with repeatable and reliably automatable tasks, and we're now prototyping more advanced use-cases with customers. For instance, one customer initially planned extensive hiring for a complex financial workflow but shifted to our agent platform after seeing how efficiently these workflows could be automated.

Our core audience is still developers, and we recommend they initially experiment with workflows that simplify code deployments, API development, or general coding productivity. Once comfortable, they can extend to more advanced or even non-technical workflows.

Additionally, AI Agent Builder is indeed an evolution of Postman Flows. Once we recognized the potential of agent-based application architecture, we enhanced the capabilities of Flows to support agent development, and deploy workflows to the cloud without any additional infrastructure.

MCP has shifted the paradigm for what it means to build AI Agents. What’s your approach to helping devs build in the MCP world?

Open protocols like MCP are essential for scaling AI agents effectively. They remove vendor lock-in, making it simple to integrate, swap, or add new capabilities and models into your applications. By standardizing how models interact securely with APIs, tools, and data sources, MCP significantly reduces integration complexity and lowers development costs. This flexibility is crucial—it allows developers to innovate rapidly while maintaining precise control over their systems.

At Postman, our MCP client takes this even further by significantly enhancing the developer experience. Think of it as a vastly improved version of your traditional MCP inspector, offering comprehensive support for tools, resources, and prompts. With MCP integrated directly into Postman, developers can work as intuitively as they do with REST, GraphQL, or gRPC APIs, all within a familiar and powerful environment.

Additionally, our MCP server generation capability—available through our AI Tool Builder—enables developers to effortlessly convert any public API into an MCP-compatible server. This server can integrate seamlessly across various agent platforms, even outside the Postman ecosystem. Soon, this powerful capability will extend to private APIs as well, further broadening the potential for innovation and integration.

There are a number of talented teams working on AI agent frameworks for developers. What sets Postman’s AI Agent Builder apart from a developer perspective?

The strength of our AI Agent Builder lies in our extensive API ecosystem. Agents can only perform as well as the APIs they can access. Postman is perfectly positioned to get developers into making and using MCP servers––we already host a comprehensive collection of internal, partner, and public APIs, positioning us uniquely to empower developers to build richer, more powerful agents without friction.

Unlike other examples on social media, real-world agent workflows typically involve combining multiple private APIs alongside public APIs. Postman provides immediate, easy access to both categories within a single, unified platform, eliminating common integration roadblocks faced elsewhere and eliminating the need to jump around tools to build their agent.

Moreover, our identity as a core developer tool allows continuous iteration and improvement based directly on extensive developer feedback. This deep developer engagement ensures a superior and constantly evolving user experience.

On a personal level, is it scary to shepherd this company, your baby for the last 13 years, into a complete transformation of the AI era?

You tend to feel fear only when you don't understand something completely. Personally, I'm a learner at heart and a technology geek, so I genuinely enjoy diving deep into generative AI. I've been trying to learn how to build an LLM, experimenting with agentic applications, and exploring various possibilities firsthand. For me, this wasn’t about jumping on a bandwagon—it aligns perfectly with our original mission at Postman, symbolized by our space-age astronaut logo, which was to elevate developer tools into the future.

Generative AI now gives us a chance to rethink foundational aspects of software. For example, I believe that about 90% of the fluff UI we currently have will simply vanish because it won't be needed anymore. The product managers of tomorrow will just focus on building robust APIs, with AI agents discovering and implementing workflows dynamically. This represents a dramatic shift from how software was built just a few years ago.

I’m extremely excited about these possibilities. APIs have always been essential and continue to grow in importance. This new world of agentic applications is built on millions and billions of APIs, and being part of this transformation is truly exhilarating.

Your prediction that the focus in product development will shift from UI/UX to API design is, in my opinion, right on the money. How does that realization affect your strategy for Postman in the next year? The next 5 years?

Over the next year, Postman will evolve into a deeply AI-native application. Instead of viewing AI as a separate add-on, we are integrating it fundamentally into every aspect of our workflows. This integration will enable collaborative interactions between humans and AI agents. For instance, a junior engineer previously might have needed to invite a senior or staff engineer to review their work. Now, that engineer can also rely on an AI agent to help review and refine their efforts, dramatically expanding their capabilities.

We are actively building MCP tools for every single feature within Postman, making them readily accessible to AI-driven workflows. This holistic approach means reimagining existing boundaries rather than working within them.

Looking five years ahead, I see agents actively participating in production environments—not just observing, but reliably performing actions on our behalf. Current challenges around agent reliability, trust, and identity will be progressively solved, leading to far more complex and trustworthy multi-agent systems. The result will be transformative, reshaping not just technology but also society's broader expectations and acceptance of AI.

What would you say is a personal quality of yours that has found its way into the culture of Postman as a whole?

I’ve always been a builder. I like to take things apart, understand how they work, and try out new ideas, especially with new technologies. Even outside of work, I’m usually tinkering or prototyping something. This same curiosity is a big part of Postman’s culture. Our weekly Demo Days are a great example of this as they give teams a space to explore, share, and turn ideas into real product improvements.

Tell us a bit about the AI organization at Postman. Are you hiring and for what roles? What is something you'd see in a prospective hire that tells you "yep they're a fit at Postman"?

We have a cross functional team that is building agentic capabilities, along with a dedicated product team for our agent platform. We are also hiring applied AI engineers on almost all teams. In fact, one of my cofounders is a key contributor to AI initiatives.

In terms of hiring, we look for people who know their craft and can plug into a team, close gaps, and keep things moving even when the path isn’t clear. This is especially important in hiring for AI roles, where everything is changing fast and we need people who can adapt and build as they go.

Conclusion

Stay up to date on the latest with Postman’s AI efforts, follow them here.

Read our past few Deep Dives below:

• Arcade - A New Kind of Marketplace 🔍

• Strong Compute - Your High-Performance Training Infra Platform 🔋

• V7 Go - AI Agents for Complex Document Workflows 🎛

• Zencoder’s Repo Grokking - Your AI coding unlock 🔑

• Rime - Authentic AI Voices for Real-time Conversations at Scale 🗣️

If you would like us to ‘Deep Dive’ a founder, team or product launch, please reply to this email (newsletter@cerebralvalley.ai) or DM us on Twitter or LinkedIn.

Join Slack | All Events | Jobs