May 2026 Summaries

Insomnia 12.6 introduces significant updates aimed at enhancing developer workflow by integrating native Git CLI support and dynamic mocking capabilities, both for self-hosted and cloud-hosted environments. This version allows developers to manage their API workspaces natively with Git, facilitating seamless collaboration and integration with existing CI/CD pipelines. The real-time auto-sync feature ensures that any changes made in the terminal or editor are instantly reflected in Insomnia, eliminating the need for manual imports or refreshes. Additionally, "Try in Insomnia" deep links provide an immediate API testing experience, reducing the steps between discovering and testing APIs to a single click. These enhancements are designed to give developers flexibility and control over their tools, workflows, and data without incurring extra costs, contrasting with competitors like Postman, which require paid plans for similar collaborative features.

Anthropic's acquisition of Stainless highlights the ongoing trend of making AI agents more effective by shortening the path from APIs to agent-ready interfaces. This move underscores the importance of connectivity and governance in building scalable agentic systems, as the ease of generating SDKs and MCP servers does not address the critical governance questions that arise during runtime. These include managing agent access, ensuring seamless operation despite changes in API rate limits, and attributing costs accurately. Furthermore, the evolution of agentic systems requires a robust runtime governance layer that is vendor-neutral and capable of handling real-time data triggers and memory continuity, rather than relying solely on build-time tools optimized by model vendors like Anthropic and OpenAI. Enterprises must focus on establishing this independent governance layer to manage security, observability, and monetization effectively across multiple models and agent frameworks, ensuring that their AI systems are reliable and business-ready.

Kong Konnect and VMware vSphere Kubernetes Service (VKS) address the challenges of scaling Kubernetes environments by providing a unified, policy-driven platform for managing APIs, AI models, and microservices with a centralized control plane. Kong's Data Plane Gateway acts as an intelligent control tower, ensuring secure and governed internal traffic, while the Kong AI Gateway enables orchestration of multiple GenAI providers, agent frameworks, and traditional REST-based APIs. This integration streamlines operations, enhances security, and offers deep insights into traffic patterns and service performance. VKS, as a CNCF-certified distribution, provides an enterprise-grade Kubernetes runtime integrated into VMware Cloud Foundation, leveraging existing infrastructure expertise to simplify the deployment and management of containerized applications. Together, Kong and VKS transform the API gateway into a strategic control plane that bridges traditional infrastructure with AI-native agility, enabling organizations to innovate while maintaining performance and security standards.

As AI infrastructure generates financial signals that often go unnoticed by CFOs, bridging the gap between technical metrics and financial reporting becomes crucial for effective decision-making. The AI connectivity platform can serve as both a technical and financial control plane, offering insights into token consumption, cache hit rates, and model routing decisions, which are vital for cost management and profitability. This guide emphasizes the importance of translating technical data into financial terms for CFOs, highlighting the need for infrastructure observability to align with financial reporting. It suggests starting conversations with CFOs by focusing on key metrics that reveal cost exposures and margin improvements, ultimately enabling informed financial decisions. The AI gateway captures essential data on workload costs and customer service costs, yet this information often resides in technical observability tools rather than financial reports. By effectively communicating the financial impact of AI operations, organizations can leverage existing systems for enhanced financial visibility, thereby aligning AI initiatives with business objectives and ensuring that finance teams understand the economic implications of AI investments.

AI systems, particularly agentic ones capable of autonomous decision-making, face challenges in traceability and accountability due to their reliance on static data snapshots, like vector databases and key-value stores, which fail to capture the sequential reasoning process. This lack of a comprehensive reasoning trace can hinder observability, compliance, and debugging. A shift towards treating the event stream as the source of truth, akin to a durable commit log, allows for a detailed, ordered record of every decision, tool call, and context shift, ensuring a system that is observable, governable, and trustworthy. This approach advocates for using event streaming systems, such as Apache Kafka, to capture and govern the reasoning trace, enabling replayability, governance, and a unified trace across all infrastructure layers. By implementing governance at the connectivity layer with tools like Kong AI Gateway and Kong Event Gateway, organizations can ensure visibility, security, and control over the entire data path, transforming AI from a black box to a transparent, accountable system capable of explaining its decision-making processes.

The Model Context Protocol (MCP) plays a crucial role in enabling AI agents to communicate with external tools and data sources by standardizing their interactions, similar to how HTTP functions for web communication. However, without an MCP registry, agent-to-tool connections risk becoming disorganized and inefficient, akin to the challenges faced by microservices before the advent of API gateways. An MCP registry serves as a centralized directory where AI agents can discover MCP servers and the tools they offer, storing metadata about these servers rather than hosting the tools themselves. This setup addresses issues like configuration drift, duplicated integrations, and ungoverned access, providing a single source of truth for tool discovery and governance, with the registry facilitating dynamic discovery and controlled access. While the registry acts as the discovery layer, an MCP gateway enforces runtime policies, ensuring secure and authorized interactions. Enterprises benefit from both the registry and gateway, achieving scalable, governed, and efficient AI agent deployments that are essential for transitioning from pilot projects to production at scale, with solutions like Kong MCP Registry offering comprehensive governance and observability features.

GitHub's recent shift from seat-based to consumption-based pricing for its Copilot service marks a significant change in how AI-driven products are monetized, highlighting a broader trend toward custom currency models in the tech industry. This new pricing strategy involves AI credits, which serve as an intermediary currency that abstracts the unit of value from the cost, allowing for greater pricing flexibility as infrastructure costs fluctuate. This model is already utilized by companies like Canva and Salesforce to manage the variable and growing costs of AI infrastructure, providing a stable pricing framework that aligns with business outcomes rather than raw token usage. The change has sparked debate, particularly among power users who feel burdened by the new structure, while lighter users may benefit from pooled credits and better financial governance. This transition underscores the importance of integrating metering and monetization strategies early in product development to avoid disruptive shifts later. The adoption of custom currency models is becoming the standard for AI products, necessitating organizations to prioritize governance and flexible pricing as foundational elements of their business strategy.

The Model Context Protocol (MCP) server, introduced by Anthropic in November 2024, serves as a lightweight intermediary that links AI applications with external systems such as databases, APIs, and SaaS products through a standardized protocol. Unlike traditional methods requiring bespoke integration code for each connection, MCP servers provide a unified interface, enhancing the ability of AI-powered applications to interact with multiple external systems efficiently. MCP servers operate by exposing capabilities like tools, resources, and prompts to AI applications, allowing for capability negotiation and standardized invocation patterns. While the protocol itself defines the rules for interaction, the server implements these rules, similar to the relationship between HTTP and web servers. This system enables AI applications to connect with various external services using a consistent, self-describing interface, thereby minimizing maintenance burdens associated with custom integration code. However, deploying MCP servers at scale introduces challenges such as authentication management and observability, which can be addressed by implementing centralized governance solutions like the Kong AI Gateway. This setup facilitates seamless, secure, and scalable interactions between AI applications and their required external systems.

A recent Gartner research note highlights the challenges enterprises face in integrating AI agents with their applications, emphasizing that traditional systems were not designed to handle the dynamic, unstructured requests typical of AI models. The difficulty lies in the need for an AI control layer that governs and provides observability, ensuring secure interaction between AI agents and enterprise systems through routing, authentication, and auditability. The report recommends a three-pillar approach for integration: developing AI-consumable interfaces, implementing an AI control layer, and ensuring agent-ready data, all to facilitate seamless and secure AI deployment. The Kong AI Gateway is highlighted as a solution that provides the necessary governance infrastructure, supporting both AI and traditional API traffic while maintaining auditability and security. The research underscores that organizations will succeed by extending existing API management systems to accommodate AI use cases, rather than replacing them, thus ensuring a robust and governed AI integration strategy.

An enterprise AI gateway serves as a centralized control plane to manage, secure, and route AI traffic at scale, with LiteLLM and Kong being prominent examples. LiteLLM is an open-source AI gateway that offers baseline functionalities such as multi-LLM routing and governance, making it suitable for initial AI connectivity needs. However, as organizations scale, Kong stands out due to its advanced enterprise features, including higher throughput, lower latency, and comprehensive governance capabilities that extend beyond basic connectivity to include agent-to-agent traffic management and centralized cost control. Kong's architecture, built on a compiled core for optimized performance, allows for more efficient handling of high-volume traffic compared to LiteLLM's Python-based proxy layer. Additionally, Kong provides enhanced security measures, such as centralized PII masking and robust access controls, which are critical for production environments. As AI platforms become integral to enterprise operations, the need for such comprehensive governance and performance capabilities becomes paramount, positioning Kong as a preferable choice for more demanding production requirements.

Apache Kafka is widely adopted due to its scalability and reliability, powering real-time systems across various industries. However, as Kafka usage scales across multiple teams and regions, it presents challenges such as inconsistent security policies, fragmented observability, and reactive governance. Traditional governance models, which rely on Kafka's "smart client" approach, often break down at scale, leading to risks such as over-permissioning and compliance issues. To address these challenges, organizations are turning to Event Gateways, which act as centralized policy enforcement points between clients and Kafka. These gateways offer fine-grained security, built-in data quality enforcement, and compliance automation, transforming Kafka into a more manageable and observable platform. By decoupling routing, security, and transformation logic from applications, Event Gateways allow developers to focus on feature building while providing platform teams with centralized governance. This approach not only enhances control and consistency but also transforms Kafka into a shared platform that meets growing compliance and security demands.

As organizations shift from experimental AI to production-grade systems, they encounter challenges such as fragmented LLM providers, complex tool integrations, and increased security risks. A new infrastructure stack is emerging, emphasizing the need for scalability and security, which Kong and Akamai address through their combined offering. This infrastructure comprises three layers: the Kong AI Gateway for managing GenAI and MCP flows, the Akamai Linode Kubernetes Engine for computing, and the Akamai Firewall for AI for security. The Kong AI Gateway provides a unified API layer, enabling applications to switch models without code rewrites, while offering features like credential management and semantic processing. Meanwhile, Akamai's Firewall inspects requests in real-time, understanding intent and safeguarding against AI-specific threats like prompt injection attacks. This partnership allows organizations to streamline AI operations, enhance security, and speed up deployment, transforming AI from a security risk into a competitive advantage.