Issue #84

AI made PMs faster. Multiplayer mode is still broken.

A PM can summarize research, draft a PRD, and mock up a prototype before lunch. The hard part starts when the team has to decide what actually gets built.

Jira Product Discovery gives product teams one place to capture insights, prioritize ideas with consistent frameworks, and build living roadmaps stakeholders can rally around.

And because it’s connected to Jira, the context behind every decision stays with the work—so developers and their agents know not just what to build, but why.

AI helps PMs move faster. Jira Product Discovery helps the whole team build with confidence.

Get Started for Free.

Welcome to another issue of the APIPlatformsForScale Newsletter. The video of my talk at APIdays India has been published, and you can view it here.

Another talk I recommend from APIdays India is Sriram Narayan’s talk on Impact Intelligence for APIs. Sriram discusses how to measure the business impact of API development.

With that said, here are some interesting links for you in this issue.

MCP just changed a fundamental part of how it works, and the implications go well beyond the protocol itself. Janet Wagner explores what the move to stateless MCP means for scalability, gateways, security, and API providers. By removing protocol-level sessions, MCP can scale more naturally across standard HTTP infrastructure—but it also shifts important responsibilities around state, identity, and recovery.

As AI accelerates API development, one problem is becoming harder to ignore: tests are struggling to keep pace with constantly changing API specifications. When the spec changes but tests remain tied to an older version, teams can end up discovering drift only after something breaks in staging. Vyshnavi Dabbir’s guide explores a spec-driven approach that connects OpenAPI definitions directly to functional testing—from automatically generating tests and managing authentication to integrating coverage into CI/CD and AI coding assistants through MCP.

As AI agents move from isolated experiments into core enterprise workflows, the governance challenge is shifting from managing individual platforms to governing an entire agent estate. BCG argues that platform-by-platform controls create fragmented identity, visibility, security, and cost management and proposes an Enterprise AI Control Plane to bring these concerns together.

As organisations move from a handful of AI experiments to hundreds of agents, tools and skills, discoverability, and governance are becoming the next major bottlenecks. AWS has now made Agent Registry generally available, providing a central catalogue for registering, discovering and governing MCP servers, A2A agents, skills and other agentic resources.

What happens when an API is judged not only by what it exposes but by the capabilities it explicitly cannot provide? This APIs.io piece looks at Archive’s API through that lens, turning limitations into useful information about the boundaries of a developer platform.

AI has made software development faster, but it has not necessarily made it easier. The Mia-Platform team shares that much of the friction has simply moved elsewhere, with developers now dealing with tool sprawl, fragmented environments, AI-generated code they cannot fully trust, and growing governance demands.

The next challenge for AI agents may not be making individual agents smarter, it may be about stopping millions of smart agents from making the same decision at the same time. MIT Media Lab argues that the internet was designed around human-scale interaction, while agents can coordinate at a density and speed that creates entirely new failure modes.

The AI infrastructure race is creating a new challenger to the traditional cloud giants: neoclouds built specifically around GPU-intensive workloads. They promise lower costs, faster provisioning and infrastructure optimised for AI training and inference, but the hyperscalers still hold significant advantages in security, governance, managed services and enterprise integration.

AI governance has largely focused on what happens in the cloud. But what happens when developers start running agents, MCP servers and AI tools directly from their laptops? Solo.io’s AgentDesktop explores a new approach to this problem, bringing visibility, security, and governance closer to where AI activity is actually happening. It raises an important question for platform teams: is the developer workstation becoming the next frontier of agent governance?

A major cryptography compliance deadline is approaching, but many engineering teams may not realise how much of their infrastructure it affects. On 22 September 2026, remaining FIPS 140-2 certificates move to the CMVP Historical List, putting greater pressure on organisations working with government and other regulated environments to move to FIPS 140-3.

Building an MCP server can still involve a surprising amount of engineering work, particularly when the goal is to expose existing APIs as agent-ready tools. Bill Doerrfeld takes a look at the growing range of MCP server generators and asks what really separates the approaches.

What happens when AI agents become capable enough that traditional sandboxes and rigid guardrails start getting in the way? Steve Yegge argues that the next generation of agentic systems may need something fundamentally different: rules that govern what agents are allowed to do, rather than walls designed to contain them.

About this newsletter

The API Platforms for Scale Newsletter curates intelligence on the trends, technologies, and practices shaping API platforms, MCP, developer, and agent experience.