AI Pulse · Under the Radar

A company that sells temporary phone numbers for SMS verification just rebuilt its entire front end for AI agents instead of people. SV Number wrapped its 200-country API in an MCP server so an agent can request a phone number, wait for the verification code, and finish a signup flow without a human touching the keyboard. The GitHub repository picked up 534 stars in three days, and the team shipped a companion “skills” repo the same week to show other developers how to wire it up.

This matters because it’s a clean example of a pattern you’ll see more often: companies pivoting from human-facing SaaS to agent-facing infrastructure. SV Number didn’t invent a new service. It repackaged an existing one so agents can call it directly, pay per use, and move on. No dashboard, no onboarding flow, no support chat. The agent is the customer.

For anyone running a niche API business, this is worth studying. If your service solves a repetitive task that agents will hit over and over, wrapping it in an MCP interface and pricing it for programmatic access could open a new revenue stream before your competitors do. The technical lift is modest. The question is whether your current product can work autonomously, because agents won’t wait for manual approvals or call your support line when something breaks.

If you’re building workflows that string together multiple third-party services, this is the kind of modular infrastructure we wire into systems like the Omni Command Centre so agents can handle end-to-end tasks without stopping for human input. The shift from human-in-the-loop to agent-direct is happening fast in narrow verticals, and the companies that move early will set the pricing and integration standards everyone else has to follow.