Kong

Take it from a Konger who recently joined the team, the people at Kong are amazing! 🦍 Check out the video to hear it directly from Abhinav Goyal, Staff Customer Success Manager. Interested in joining a category-defining company with a values-driven culture and amazing colleagues around the world? Check out our open roles below. 👇 #LifeAtKong

I remember when Reza Shafii first started talking about how "Integrations are dead (link to the original post in the comments)." Took me a bit to be totally convinced of the claim, and--yes--we of course don't actually mean that all integration efforts are totally dead. But the larger market shift towards Agentic AI has eventually convinced me that the legacy approach to integration is dying. And now here comes a new Gartner report also talking about the need for a new approach to integration, focused on a new idea around a "Context Mesh" that posits a new AI + API based approach to agentic integration. Turns out, the initial hot takes from some of the folks here at Kong were right on the money. More details on the Gartner report and what it means for API and AI programs in the comments below.

🤖 AI should be a strategic lever for speed, cost control, and competitive advantage. But fragmented infrastructure is getting in the way. On February 2nd, live from the NYSE, Kong leaders will share how enterprises are unifying their AI connectivity to move from pilot to production — fast. 🏎️ You won't want to miss this. Register for the livestream here > https://bit.ly/4jILIBd

Modern AI agents need context. And context is delivered via tools–often through MCP. But how do you enable secure, governed access to MCP and tools? With MCP Tool Access Control Lists (ACLs), organizations can implement granular authorization policies that control which tools within an MCP server can be accessed by specific users, applications, or AI agents. 📣 And now, MCP Tool ACLs are available in Kong AI Gateway 3.13! Enterprises can filter tools based on identity, implement default-deny policies, leverage consumer groups, and maintain RESTful upstream APIs. 👉🏼 Check out our blog for more info: https://bit.ly/4qQNKBH

This architectural shift gives customers control over their data and costs while vendors focus on innovation. A win-win for cloud software of tomorrow. By Hugo Guerrero, thanks to Kong

🗣️ "We’re developing products and features faster than any company that I’ve ever been a part of. And we’re doing that so we can take advantage of this massive AI market that’s really exploding. It’s creating this huge tailwind for Kong that as a seller at Kong, and just an employee at the company, it makes it a really exciting place to be.” Andrew Balter, Enterprise Account Executive Interested in joining the company building the connectivity layer of AI? Check out our open sales roles: https://bit.ly/4jzEXla #LifeAtKong #KongSalesTeam

Most companies have zero idea what their AI is actually costing them. I’m talking to CFOs who are terrified by this. They greenlit a pilot, it worked great, and then they tried to scale it. Suddenly, they are spending more on API calls and token consumption in a single month than they budgeted for the entire quarter. Despite all the attention and hype, most companies have no visibility into their AI data paths. They can't see which model calls are driving value and which are just burning money. And cost only one part of the problem. When I talk to engineering leaders, the blockers are always the same: speed, cost, and risk. You cannot solve these problems with a patchwork of different tools. You need a unified approach. On Feb 2nd at the NYSE in New York City, we’re defining the path to address these three blockers. Stay tuned: https://lnkd.in/eNQY2WGa

𝗔𝗿𝗲 𝗮𝗴𝗲𝗻𝘁𝘀 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝘀𝗼𝗺𝗲𝘁𝗵𝗶𝗻𝗴 𝗻𝗲𝘄? Or just smarter applications with better marketing? This is a question that has been bothering me for a while. For a long time, I resisted the term “agent”. I kept asking myself why not just say LLM-powered apps? Over the course of last year I’ve come to peace with the term agent, because I think it reflects a real shift in how we design systems. Here’s where I landed. 1️⃣ 𝗨𝗜 𝗹𝗼𝗴𝗶𝗰 𝘁𝗵𝗶𝗻𝘀 𝗼𝘂𝘁 Traditional apps optimize UI for human interaction. Agentic systems optimize for helping humans leverage LLMs effectively for the task at hand. That changes the job of the UI: • Less about encoding business logic into flows and screens • More about deciding how much autonomy to give the model vs how much control to keep with the user I don’t believe all UI disappears behind a single prompt box. But UI will increasingly exist to shape, validate, and scope what gets handed to an LLM. Cognitive load shifts away from memorized workflows toward helping models get the right information or perform the right task. 2️⃣ 𝗕𝗮𝗰𝗸𝗲𝗻𝗱 𝗹𝗼𝗴𝗶𝗰 𝘀𝗵𝗶𝗳𝘁𝘀 In classic architectures, backend logic drives behavior through explicit rules and workflows. In agentic systems, behavior emerges from how context, tools, and constraints are assembled at runtime. More “thinking” moves to the LLM, while the backend focuses on: • Retrieving the right context from the right APIs and data sources • Deciding what to pass to the model and when This is where MCP comes in — as a mediation layer for assembling the right context so LLMs can do what they’re good at. 3️⃣ 𝗔𝗣𝗜𝘀 𝗯𝗲𝗰𝗼𝗺𝗲 𝗲𝘃𝗲𝗻 𝗺𝗼𝗿𝗲 𝗰𝗿𝗶𝘁𝗶𝗰𝗮𝗹 APIs were already critical infrastructure. Agents raise the bar because, in agentic systems, the thin front-end layer is primarily about orchestrating LLM interactions and retrieving context from the right API tools, while the backend increasingly focuses on exposing and governing those tools in a way that optimizes the agent’s purpose. As a result, APIs stop being passive integration points and become inputs to reasoning. Bad APIs won’t just fail — they will mislead models, and the cost of that will become increasingly obvious. 𝗦𝗼… 𝗮𝗿𝗲 𝗮𝗴𝗲𝗻𝘁𝘀 𝗷𝘂𝘀𝘁 𝘀𝗺𝗮𝗿𝘁𝗲𝗿 𝗮𝗽𝗽𝘀? At this point, I don’t think so. Taken together, the changes in the nature of the beast are finally worth the new name in my mind. #AgenticSystems #AIArchitecture #LLM #APIs #ContextEngineering #PlatformEngineering

My 2026 hot take: speed isn’t everything. I’m already seeing some of our customers make the shift to prioritize governance-by-default and AI traffic observability rather than speed for the sake of speed. Mostly when testing turns to scale and production. The next step for AI isn't just about shipping faster, it's about running safer and with costs under control. With the proliferation of agents, APIs become the universal language connecting models and microservices. Without a unified API layer, you’ll just end up sprinting towards tool sprawl and added risk. You've invested in Platform Engineering. I am sure You've bet on AI. Now, build with intention.