Fewer than one in twenty MCP servers currently charge money for access. The other nineteen are free — built by developers scratching their own itch, released as open source, or deployed as demos that were never intended to become products.

This ratio makes sense as a description of where the ecosystem is right now. MCP as a protocol is still relatively young, and the developer community that built the first wave of MCP servers built them the way developers typically build things: to solve their own problems, with the tools and workflows they understand, and without a strong commercial model in mind. The result is a large and growing catalog of free tools, most of which are useful but none of which are businesses.

The five percent that do charge represent something different. They’re typically tools built for problems that developers didn’t personally have — professional workflows, domain-specific needs, data access that requires ongoing maintenance and licensing costs. These tools are harder to build, require domain expertise that isn’t widely distributed, and serve buyers who are accustomed to paying for professional software. The commercial logic exists from the beginning because the problem couldn’t be solved by someone building for free.

The monetization gap in the MCP ecosystem is therefore also a description of where value is accumulating. Free tools built by developers for developers are commoditizing quickly — there are often multiple free alternatives for any given developer workflow, and the marginal value of one more is approaching zero. Paid tools built for non-developer professional workflows don’t have that problem. There are almost no alternatives, free or paid.

This creates a straightforward opportunity structure: the MCP ecosystem is undersupplied with domain-specific professional tools, and the tools that do exist in this category face very little competition. The professionals who need them are accustomed to paying for professional software. The reference price point for a professional AI tool is not the average MCP server — it’s the existing software budget those professionals already have.

The five percent is not a ceiling. It’s a starting point that reflects where the ecosystem’s builders have focused attention so far, which is not where most of the unmet need exists.

+++