Amsterdam was a good one for NitroStack.
On September 17 and 18, 2026, AGNTCon + MCPCon Europe brought the MCP and agentic AI ecosystem together at RAI Amsterdam.
We spent those two days talking with founders, developers, infrastructure teams and people building different parts of the agent stack.
A lot of those conversations came back to the same question:
What does it actually take to move MCP from a working demo into something teams can use in production?
That is exactly the part of MCP we have been building NitroStack around.
The MCP conversation has changed
A few months ago, a lot of MCP conversations still started with explaining what the protocol was.
Amsterdam felt different.
The sessions and hallway conversations were much deeper into stateless MCP, authentication, agent identity, security, MCP Apps, governance, observability and production infrastructure.
That is a healthy sign for the ecosystem.
People are starting to care less about whether they can connect one tool to an agent and more about what happens after that tool becomes part of a real product.
How does it scale?
How do you authenticate users?
How do you understand what the agent actually called?
How do you deploy the server cleanly?
What happens when the workload grows?
These are already very familiar questions inside NitroStack.
A lot of this already exists inside NitroStack
NitroStack is being built as an end-to-end development stack for Model Context Protocol applications.
The open-source NitroStack SDK gives developers the application framework.
NitroStudio gives them a place to build, inspect and test MCP tools, resources, prompts and workflows.
NitroCloud handles the deployment and infrastructure side.
NitroChat can take that MCP server and turn it into something people can actually use.
For us, these are not separate products solving unrelated problems.
They follow the same developer journey:
Build → Test → Deploy → Operate → Use
That full path became especially relevant in Amsterdam because so many conversations were already moving toward production MCP.
Stateless MCP was an important part of the discussion
Stateless MCP came up repeatedly during the event.
This is an area NitroStack has already spent a lot of time working on.
We recently published Introducing Stateless MCP to NitroStack, where we explained how moving away from server-held sessions simplifies the infrastructure behind an MCP application.
A request no longer has to return to the same server instance.
That makes horizontal scaling cleaner and makes MCP behave much more naturally with modern cloud infrastructure.
Once applications start receiving real traffic, details like this stop being implementation trivia.
They directly affect how reliably the system can run.
This is one of the reasons the technical direction of MCPCon Europe felt so close to the work happening inside NitroStack.
Security and identity are becoming part of the MCP foundation
Another big topic was security.
That makes sense.
An MCP server may eventually be able to read internal data, update a CRM, trigger a workflow, create a ticket, make a purchase or interact with sensitive enterprise systems.
At that point, authentication cannot be added at the end.
The server needs to understand who is making the request, what that user is allowed to access and what the agent should be allowed to do on their behalf.
NitroStack has been building these controls into the MCP application layer itself, alongside validation, middleware, authorization and production deployment.
For enterprise MCP, this is going to matter a lot.
The most useful part of Amsterdam happened between sessions
The technical program was strong, but a lot of the value for NitroStack came from conversations outside the sessions.
We met teams building across developer tooling, infrastructure, security, AI products and the MCP ecosystem.
Quite a few conversations turned into potential collaboration paths.
In some cases, a company already had a product or API that could become much more useful when exposed properly through MCP.
In others, the team was building one part of the agent stack and NitroStack could provide the development or infrastructure layer around it.
There were also conversations around integrations and partnerships that we are continuing after the event.
Those discussions are still early, so there is not much value in naming things before they are real.
But the fit was clear in several places.
NitroStack is becoming easier to explain
One thing we noticed in Amsterdam was that the NitroStack story has become much simpler.
Earlier, MCP itself needed a lot of explanation.
Now people increasingly understand the protocol.
The next question is usually:
How do I actually build and ship with it?
That is where NitroStack becomes easy to understand.
You build the MCP application with the SDK.
You work with it inside NitroStudio.
You deploy it through NitroCloud.
You can take it into an end-user experience through NitroChat.
The infrastructure underneath can deal with things like scaling, authentication, observability and deployment while the developer focuses on what the MCP application actually does.
That is the layer we want NitroStack to own.
A strong two days in Amsterdam
AGNTCon + MCPCon Europe was valuable for NitroStack because the ecosystem is now discussing many of the same problems we have already been building around.
Production MCP.
Stateless infrastructure.
Security.
Agent identity.
Deployment.
Developer experience.
And how MCP applications become real products rather than experiments.
Amsterdam also opened several conversations around partnerships and integrations that we are excited to continue.
For us, that was the best outcome from the event.
The MCP ecosystem is moving quickly.
NitroStack is right inside that movement, building the infrastructure developers need to actually ship with it.