AI agents can tell you what they know. But can they tell you how they got there? ๐ค
AI observability needs needs more than prompts, tokens and model responses. Developers need a complete, replayable history of every decision, tool call, and context change.
Check out this blog from @hguerreroo on why durable commit logs may be the missing foundation for AI observability ๐๐ผ https://t.co/W7Gwq1YriQ
Every company in the AI era is going to face the same structural question: what do you actually own?
Models are commoditizing and agents are making interfaces irrelevant.
Your context layer is your identity, but most enterprises don't even know they have one.
Between the systems of record that hold your data and the agents that act on it, there's a layer that doesn't have a name in most org charts. It's the semantic intelligence wrapped around your data โ the context that tells an agent not just what the data is, but what it means, when it's relevant, and how to use it.
Context is what agents eat. Via APIs, via MCP tools, via event streams flowing across your organization.
The richest context in any enterprise already exists, buried in API traffic, transactional systems, and years of operational signals.
So many people in tech confuse "API requests" and "tool calls."
Tool call = standardized intent
API request = programmatic execution
MCP = the protocol for agent-to-tool traffic
A2A = the protocol for agent-to-agent traffic
An API request is software calling software. The plumbing that's been running the internet for 30 years.
A tool call is the AI's decision: when an agent needs to book a flight or pull a record, it doesn't write HTTP. It issues a structured instruction, and the system figures out the rest. One tool call can trigger five or six underlying API requests.
MCP sits between them โ tool discovery, context passing, structured results. It's what makes tool calls portable across models and platforms.
But agents don't just call tools. Agents call agents.
That's where A2A comes in. MCP governs the conversation between an agent and a resource. A2A governs the conversation between two agents. Different trust model. Different authorization chain. Different governance problem entirely.
Securing a tool call is not securing an API request. Metering MCP is not metering REST. Governing AI traffic means governing the whole stack.
Gartner Application Summit day 1 โ
It was a day full of conversations about how Kong helps leading teams govern AI without slowing developers down, monetize AI-driven architectures, manage API, event and AI traffic with a single control plane, and much more.
There's still time to see us this week! Come by booth #207, join our happy hour tonight, or book time with our execs ๐๐ผ https://t.co/InaYYWUHlQ
I used to think in years. Now I think in decades.
When I started @Kong, I might have guessed we'd hit certain milestones sooner than we did. But looking back, we've also built something FAR bigger than anything I could have mapped out at the start.
Both things are true at the same time โ and that tension taught me something I wish I'd understood earlier.
We overestimate what we can do in the short term, and we dramatically underestimate what we can do in the long term.
The companies that compound โ the ones that end up defining their era โ aren't necessarily the ones moving fastest in year one. They're the ones who identified something foundational.
For us, that was APIs. The world was going to generate more APIs, more connections, more traffic โ no matter what. That was true in 2015. It was true in 2020. It's more true today than ever, now that every AI agent on the planet runs on APIs. We didn't need to predict the LLM revolution. We just stayed close to something fundamental and made sure we did it better than anyone else in the world.
That's what a long time horizon gives you. Not certainty about the destination โ but the patience to let the right bets compound.
I think about this a lot as we build Kong into the connectivity layer for the AI era. We're in the middle of something that will take another decade to fully emerge, even as it changes every day. That's exactly where we want to be.
Two of our board members made the Forbes Midas List 2026! ๐
Congratulations to @mavolpi and @martin_casado! We're lucky to have you with us on our journey to build the AI connectivity layer.
Venture capital is on the cusp of what could be its greatest returns ever.
This yearโs Forbes Midas Listโthe 25th anniversary editionโreflects the staggering rise of SpaceX, OpenAI and Anthropic, three private companies whose valuations are reshaping the venture landscape and driving a record 25 newcomers onto the ranking.
Read more: https://t.co/6s7cg0qAD2 (Photos: Guerin Blask for Forbes) #ForbesMidas
Tune in to @MTSlive this afternoon to see some familiar faces. ๐ ๐ฆ
Our co-founders, @sonicaghi and @subnetmarco, will be live at 2:00pm PT/5:00pm ET
Attending Gartner Application Summit? Don't miss our session on Tuesday at 11:45am! ๐ค
Our Senior Technical Product Marketing Manager, Dan Temkin, will cover where AI cost and AI governance break down, how teams manage token usage at scale, and what sustainable AI infrastructure looks like.
Will we see you there? https://t.co/unPmpaRscc
๐ฃ Insomnia 12.6 is live! What's new?
๐น Go deep in git from your terminal with native git CLI support or stick to the Insomnia UI. Your API workspace, your call.
๐น Dynamic mocking on self-hosted + cloud-hosted mock servers
Learn more on our blog: https://t.co/7dusVqDMMU
๐ถ It's the final countdown...for our API + AI Summit Call for Papers!
If you've built something interesting with APIs and AI, we want to know your real-world stories.
The best talks come from people who have actually done the work. That means you.
Submit by Wednesday, June 10th, for the chance to present on stage in LA. Details here: https://t.co/55jH3jWMXb
Spotted in the wild: Iceland edition ๐
Proof that Kong's reach knows no limits!
๐ท Photo credit and a major shoutout to one of our customer champions whose son took their Kong water bottle on a backpacking trip through Iceland!
Did you catch our CEO and Co-Founder, @sonicaghi, on @MTSlive this week?
Watch the clip below to hear Aghi's take on Anthropic acquiring Stainless.
Then, check out the full conversation to hear about why connectivity infrastructure matters as much as models, how the internet is shifting from pixels to APIs, and how enterprises will govern AI at the traffic layer. https://t.co/km66cY4xPZ
SITUATION EXPLAINED: Anthropic bought Stainless because agents need rails.
Kong CEO @sonicaghi says the deal is about more than making Claude Code easier to use. Agents need SDKs, CLIs, MCP, and APIs to reach the systems where work actually happens.
"The connectivity infrastructure matters as much as the model... [it] is going to be the digital railroads on how AI moves."
"We're moving from pixels to APIs."
Had a great time on @MTSlive today. Great chat @theojaffee
Kicked off with a convo about Anthropic acquiring Stainless and what that means for the state of AI connectivity. Other highlights:
The headless revolution is happening in real time โ Salesforce going headless is a big deal.
The traffic layer is where AI governance has to live. Not the model. Not the data platform.
Tokens per watt and tokens per second are the metrics that will define the AI economy.
Google is processing quadrillions of tokens a month. Most infrastructure wasn't built for that.
Thanks to the MTS team for having me. Go check out the replay if you missed it.
๐ฅ Introducing our newest partner: Unfold!
Together, Kong and Unfold will accelerate channel growth across Australia and New Zealand by looking for opportunities to further engage and support partners. Unfold will help Kong's partners uncover consistent revenue streams by aligning their strategic priorities with the commercial drivers that motivate them. ๐ฆ๐บ๐ณ๐ฟ
Learn more here: https://t.co/StiaM9YoR1
AI is moving into production faster than most teams expected. Infrastructure is still catching up. ๐๐ฝ
Across enterprises, APIs, events, and AI workloads are being managed across disconnected systems โ creating gaps in governance, visibility, and cost control.
At the Gartner Application Summit, weโll be sharing how leading teams are bringing API + AI connectivity together in a single, unified platform.
If youโre attending, weโd love to connect! https://t.co/mxtU1sOuf9
๐ฃ Announcing the availability of Dedicated Cloud Gateways and the regional hosting of the Kong Konnect control plane in Singapore!
Teams now have the flexibility to run and scale APIs and AI workloads closer to where business happens while meeting security, regulatory, and data governance requirements. ๐ธ๐ฌ
To further support customers, partners, and developers, we continue to invest in Singapore as a strategic hub for Asia Pacific operations, growing the go-to-market and site reliability engineering teams there and strengthening local leadership.
Check out our press release for more info: https://t.co/vTie4O2spy
Are you an engineering leader tying AI usage back to business value? ๐ฐ
You need to be able to talk to your CFO about the metrics coming out of your AI infrastructure. And you need to do it in a way that will keep their attention, not lose their interest.
Check out our new blog for a starter guide on how to bridge engineering observability with financial reporting: https://t.co/iy5osSMKac
The @AnthropicAI acquisition of @Stainless shows where the game is heading.
It's about moving AI. Moving the context layer.
Having a strong ai connectivity infrastructure is as important as having strong models. The LLMs are limited in actions without being highly connected to the outer world. That connectivity is built and governed via APIs, CLI, SDK and MCP servers.
Whoever has the most fluid "Token path" will be able to build far superior inteligence.
Three-quarters of enterprises are piloting or deploying AI agents. The problem? Actually integrating them with enterprise applications is proving to be one of the hardest parts. ๐
Recent Gartnerยฎ research, How to Integrate AI Agents With Your Enterprise Applications, highlights the contrast between traditional integration and what AI agents require โ an AI control layer.
Kong AI Gateway is purpose-built to be the AI control layer enterprises need.
Learn more: https://t.co/mX98YOYf6d