Is MCP a sign of the reopening of the internet?

A long time ago, we had Web 2.0! Initially, this was mainly the introduction of “AJAX” - specifically the XMLHttpRequest object that allowed an already loaded webpage to make more HTTP requests to the server, making it possible to save state or retrieve more information without reloading the entire page. (In retrospect it’s pretty crazy that we had over 10 years of browsers without XMLHttpRequest …) This allowed a new generation of interactive webpages, what we today would call Single Page Applications, people were running around updating their DOM with jquery in response to XML (:sob:) fetched from the server.

Once you have the ability to load more data, people started using the same technology to not just load from YOUR server, but load data from elsewhere - and Web Mashups were born. You could freely mix and match content from different sites out there - and kinda building on both the spirit and the technology of RSS, people and products made this data freely available. For example, you could use google maps, but pull geo-coded images in from Flickr and display them on your map. Or if you had a hashtag you could pull in bookmarks from del.icio.us, tweets from Twitter, etc. Essentially, any shared indexing scheme you could make up (location, hashtags, urls, …) would let you connect your information into a web of support stuff. And it was a really nice, collaborative time!

Then Mark Zuckerberg came, and ruined it all - he realised that he could use this data to let people import their friends network from MySpace, giving people an easy way to migrate to facebook. And then he closed the gate - you could not easily get your information out of Facebook again. The same started happening everywhere, APIs closed, access became restricted and there were walled gardens everywhere.

Now - why do I write about all this in 2026? (I promise it’s only partially because I’m old and grumpy and like to sit on the porch and wistfully look at the horizon while saying “it was all so much better before”)

I think the pendulum has started to swing the other way - away from the walled gardens. With AI agents as the new and coming nexus for all the things you do, there is a new expectation that a product you use will have APIs/MCP servers and/agent skills available, so that whatever that product does can be accessed and modified by the agent.

For example, at BB we use HiBob and Float to manage holidays/timetracking - both of which now offer MCP servers (HiBob/Float). For design tools - there is a quiet revolution going on - Google launched Stitch, there is paper.design which is explicitly made for agent-human-cowork and of course Figma has their own.

It really feels like today, a product NOT having an API/skill/MCP server available - would be a good reason to not choose that product, and I have a tiny hope that we’re actually seeing the end of the walled gardens.

Which leads me to another tiny piece of hopefulness - apart from the lock-in, another reason making your information easily accessible fell out of fashion was advertising. You cannot force banner ads on people who read your posts in their RSS reader. However, making the entire internet ad-financed hasn’t turned out so well. Optimising for ad views means optimising for unhealthy consumption and rage-baiting … maybe, just maybe, we can see the beginning of the end there as well.