Do we even need to build new features any more?

Do check out what Leonid Bashlykov and the team at Revolut are doing with their Revolut X MCP offering, currently in beta.

It's a brilliant move. Technically and theoretically, I can programmatically control my Revolut account via my AI agents. The headline use case is effortlessly trading crypto from your account, but it is obviously not limited to that.

The mind boggles with the possibilities. I really like Leonid's question on his original MCP post, along the lines of: "do we even need to bother building new features any more?"

His point being that, technically, if they give me the tools via MCP or an API, then I'll do my own feature development on top of this, on an ongoing basis.

Or more precisely, my agents will.

That is a genuinely different way to think about a product roadmap. For years the answer to "our customers want X" has been to design it, build it, test it, ship it, and hope adoption justifies the cost. If the platform is properly exposed, a meaningful slice of that long tail gets built by the customers themselves, or by the software working on their behalf, and the product team's job shifts to the primitives rather than the features.

Meanwhile, I reckon I probably need a dedicated finance agent. Mine has already sorted out every single subscription I have, corporate and personal, and has the dates locked ready for negotiation ahead of renewal.