The AI stack for the agentic era.

Computing is changing, so is Postman.

Explore

From the CEO

Computing is undergoing a dramatic shift. Software is no longer primarily created, consumed, and coordinated by humans. Increasingly, software discovers, orchestrates and interacts with other software, while humans provide direction and judgment. This new type of software artifact is an AI agent.

Agents are proliferating across every organization and they are reshaping how software gets built and used, how industries function, and how work gets measured.

Agents get their abilities through APIs. Agents can gather data, take actions, retain memory, or interact with humans or other machines or agents through APIs. As they get more capable, agents are changing our relationship with computing.

Agents hold enormous power and it’s only going to grow. That power carries risk:

If you do not control your agents, your agents control your company.

Organizations are facing multiple demands at once. They need to determine:

  • Which agents they build and maintain
  • Which agents they hire from outside
  • What agents can see and do inside their walls
  • Which data and APIs agents have access to inside the organization
  • Which APIs they expose to agents outside the organization
  • Who is powering your AI: is it rented or trained in-house?

Existing approaches are insufficient

The fundamental question every enterprise is facing is no longer “How do I deploy agents?”. The challenge lies in being truly prepared for this new era and that involves answering:

How do I trust the agents I have?

Who is using these agents? When are agents hallucinating? Who are they aligned with?

If you do not know this, you are not in control of decisions.

How do I control what they can discover and access?

Have I made my APIs discoverable and accessible for agents? Have I given them the right access?

If you do not know this, you are not in control of actions.

How do I govern them?

What happens when something goes wrong? Who holds accountability? Are they working for my organization or on behalf of an outside entity?

If you do not know this, you are not in control of risk.

How do I know the ROI of my agents?

How much am I spending and am I seeing the outcomes?

If you do not know this, you are not in control of your budget.

How do I manage thousands of agents operating across the enterprise?

How do I scale my agentic enterprise? What will break if and when we start to scale?

If you do not know this, you are not in control of operations.

We are uniquely positioned to solve it

For more than a decade, Postman has helped more than 500,000 organizations discover, govern, and connect APIs. As agents become the primary consumers of APIs, that same foundation naturally evolves into the control plane for the agent economy.

We believe enterprise computing now requires a new infrastructure layer for the agent economy.

We have been building on several fronts, and together they form Postman’s Agent Stack:

  • Passport: access management for humans and agents.
  • Fabric: an AI-native gateway with support for AI, MCP, and APIs.
  • Astro AI: an agent operating system to run and control agents.
  • Fern: agent and developer experience, covering docs, SDKs, and CLIs.
  • Postman Platform: the foundation that ties together the API Catalog, the Context Graph, and the AI Engineer.

Agents have made API-first a mandate. Earlier this year we rebuilt Postman as an AI-native API platform. We shipped the API Catalog, a system of record for APIs and services, and the AI Engineer, an autonomous engineer that runs on the Context Graph sitting on top of the Catalog.

Postman’s new AI-native platform is the foundational system for these new products. The platform is the supplier of APIs and API context for Fern, Fabric, Passport, and Astro AI.

We built the original Postman for our own needs, and we built the Agent Stack the same way — through our own transformation into an agent-native company. Fabric is where we run our AI workloads. Astro AI is where we run our agents. Passport is how we give agents and humans the right access to our APIs. Fern is how we publish our documentation.

The Agent Stack is how we are accelerating in the age of AI, while staying in control.

What is Postman.ai

We are launching Postman.ai as the central place to communicate our vision and views on AI, our research, and industry benchmarking.

I am excited about the future where humans are the winners in this new era of computing.

Abhinav Asthana

Postman CEO & Co-founder

Products in the AI Stack

We are building the infrastructure for agents and APIs. Each product serves a specific layer for our customers to achieve AI sovereignty.

Passport

ACCESS MANAGEMENT

Access management for humans and agents, controlling who and what can reach your APIs across the stack.

LEARN MORE →

Fabric

AGENT-NATIVE GATEWAY

The AI-native gateway with support for AI, MCP, and APIs, governing policy, identity, and scope at the edge of every enterprise environment.

LEARN MORE →

Astro

AGENT OPERATING SYSTEM

The agent operating system that runs and controls AI agents in production, with runtime, observability, and guardrails at scale.

LEARN MORE →

Fern

DOCS, SDKS & CLIS

Agent and developer experience covering docs, SDKs, and CLIs, so every API is discoverable, typed, and callable out of the box.

LEARN MORE →

Postman

THE AI-NATIVE API PLATFORM

The foundation platform every product above is integrated with. Powered by the Context Graph, AI Engineer, and the API Catalog.

LEARN MORE →

Research from the Postman AI team.

We publish what we learn building agent infrastructure: benchmarks, evaluations, and field data on how AI agents behave against real APIs.

APIFlow-Bench: The Enterprise Survival Test for AI Agents.

A vendor-neutral benchmark measuring whether AI agents can reliably execute multi-step API workflows in enterprise contexts, run across 19 models by the Postman AI team.

AxisWhat it tests
authenticationrefresh / scope a token mid-session
discoverypick the right endpoint among look-alikes
schemarepair a body the API rejects
multistepcarry state across a chain of calls
error_recoveryback off and retry correctly
paginationwalk to the last page
statefulnessconfirm a side-effect actually landed
READ THE BENCHMARK
AGENT UNDERSTANDING88 · 78AGENT USABILITY90 · 96DEVELOPER EXPERIENCE85 · 100API DESIGN100 · 94AGENT DISCOVERY100 · 78
Dwolla — 93 · A+Twilio — 89 · A

API Scoring Portal: Are Your APIs Ready for AI Agents?

Postman scores how well AI agents like Claude, ChatGPT, and Cursor can find, read, and build with your APIs. Compare any two reports above, then see where you rank.

Score your API

The latest from across the stack.

Product milestones, partnerships, and research from the teams building the Postman AI stack.

RESEARCH

AI-Ready APIs: Agents Are the New Audience

How AI agents evaluate and choose APIs differently than developers, and what it takes to make an API ready for them.

READ →
PRODUCT MILESTONE

AI Agent Security: How Postman's AI Engineer Is Built

Inside the security measures behind the AI Engineer, including sandboxing and credential isolation.

READ →
PRODUCT MILESTONE

Our largest docs sites now render ~6.4x faster

Fern rebuilt its documentation storage from one monolithic JSON blob into content-addressed pieces, cutting load and publish times on the biggest sites.

READ →
RESEARCH

How We Really Build Production-Grade AI Agents

Agent performance is not primarily a function of the model. It rests on three coupled systems: data reliability, API consistency, and execution governance.

READ →
PRODUCT MILESTONE

Announcing the Postman Plugin for Claude Code

One install puts your workspaces inside Claude Code: mock servers, spec syncing, test runs, and AI-readiness checks, all in natural language.

READ →
RESEARCH

The Agentic OS: a first-principles approach to building agents that get things done

Why capable agents need a layered architecture — knowledge, intelligence, and strategic context — rather than a collection of independent agents.

READ →