Clearly the protocol has found PMF (protocol market fit) - that magical state where developers actually want to use something rather than being forced to
Great write up @Charlie Guo … but on the PMF as winning part, I dont know man.
I think we should all be cautious about interpreting excitement (which is what GH stars is) with AI and AI tools. Especially in the last 12-24months - there has been a influx of tinkers, who may not particularly be system builders. Its one thing to have 1000s of devs installing mcp servers in cursor. Its another to have a scalable protocol that becomes the fiber for systems built in the wild.
Thanks Victor! Fair points - I do say that MCP still has to stick the landing to become a proper long-term, scalable protocol. But I'd say it seems like it's on track to become a fixture of the AI ecosystem, especially if competitor integration is taken seriously.
One parallel I think about sometimes is the frontend framework boom of around a decade ago. There were a ton of options floating around: Backbone, Ember, React, Angular, Vue. As of 2024, those last three comprise the vast majority of usage, and two of the three come from Meta/Google. We could see a two or three-way split for different agentic frameworks, but I do think MCP is well-positioned to at least make it into the top tier.
True . There will be many experiments. MCP, NANDA, ACP,A2A, … we will all learn from them as a field and converge.
I remember r ember.js … and all the other .js frameworks. Maintaining a framework is an insane amount of work (including the openness to complete rewrites when it becomes needed?
The MCP discussion really resonates with my experience in the e-commerce integration world! That "tyranny of copy-paste" perfectly describes what we've been dealing with when connecting systems - whether it's product data flowing between platforms or trying to make AI actually useful in existing workflows.
What excites me most about the MCP approach is how it solves the exact problem I've seen with every digital transformation project: getting systems to talk to each other WITHOUT endless custom integrations. I've been experimenting with Claude Desktop's MCP capabilities for processing e-commerce data files, and while it's definitely developer-focused right now, the potential is massive. I just explored this same pattern in my latest piece on creating specialized digital solutions that don't require massive development resources: https://thoughts.jock.pl/p/build-internal-digital-solutions-fast-no-coding-required
Clearly the protocol has found PMF (protocol market fit) - that magical state where developers actually want to use something rather than being forced to
Great write up @Charlie Guo … but on the PMF as winning part, I dont know man.
I think we should all be cautious about interpreting excitement (which is what GH stars is) with AI and AI tools. Especially in the last 12-24months - there has been a influx of tinkers, who may not particularly be system builders. Its one thing to have 1000s of devs installing mcp servers in cursor. Its another to have a scalable protocol that becomes the fiber for systems built in the wild.
Time will tell I guess.
Overall, nice article
Thanks Victor! Fair points - I do say that MCP still has to stick the landing to become a proper long-term, scalable protocol. But I'd say it seems like it's on track to become a fixture of the AI ecosystem, especially if competitor integration is taken seriously.
One parallel I think about sometimes is the frontend framework boom of around a decade ago. There were a ton of options floating around: Backbone, Ember, React, Angular, Vue. As of 2024, those last three comprise the vast majority of usage, and two of the three come from Meta/Google. We could see a two or three-way split for different agentic frameworks, but I do think MCP is well-positioned to at least make it into the top tier.
True . There will be many experiments. MCP, NANDA, ACP,A2A, … we will all learn from them as a field and converge.
I remember r ember.js … and all the other .js frameworks. Maintaining a framework is an insane amount of work (including the openness to complete rewrites when it becomes needed?
The MCP discussion really resonates with my experience in the e-commerce integration world! That "tyranny of copy-paste" perfectly describes what we've been dealing with when connecting systems - whether it's product data flowing between platforms or trying to make AI actually useful in existing workflows.
What excites me most about the MCP approach is how it solves the exact problem I've seen with every digital transformation project: getting systems to talk to each other WITHOUT endless custom integrations. I've been experimenting with Claude Desktop's MCP capabilities for processing e-commerce data files, and while it's definitely developer-focused right now, the potential is massive. I just explored this same pattern in my latest piece on creating specialized digital solutions that don't require massive development resources: https://thoughts.jock.pl/p/build-internal-digital-solutions-fast-no-coding-required