Source: EIN Presswire — 2026-07-09
Summary
PatientGenie, which runs voice- and SMS-based AI agents that handle patient outreach and scheduling for health plans and health systems, connected those agents directly to RevSpring's National Provider Directory through RevSpring's MCP (Model Context Protocol) server. That gives PatientGenie's agents live access to provider phone numbers, locations, sub-specialties, clinical expertise, and availability, and in many cases the agent calls the provider's office directly by phone to book the appointment on the member's behalf — a real production MCP integration rather than an internal chatbot demo. The deal also runs the other way: through the same MCP link, RevSpring can now offer PatientGenie's scheduling agents to its own client base.
Key Takeaways
- The mechanism: RevSpring exposes its National Provider Directory through an MCP server — a standardized interface for connecting an AI agent to a live enterprise data source — instead of PatientGenie ingesting a static or periodically-synced data export.
- What the agent gets from the directory in real time: provider phone numbers, locations, sub-specialties, clinical expertise, and current availability — the actual fields needed to decide which provider fits a given member's need and whether they can be booked.
- The agent doesn't stop at lookup: PatientGenie's agents operate over voice and SMS, and for providers without digital self-scheduling, the agent places an outbound phone call to the provider's office itself to complete the booking — a genuine agent-takes-action step, not just information retrieval.
- The integration is bidirectional in business terms: the same MCP connection lets RevSpring resell PatientGenie's scheduling agents to its own client base, extending automated scheduling coverage to providers who don't support digital self-scheduling at all.
- This is a concrete, named production use of MCP in healthcare — agent to MCP server to live directory to a real-world action (a phone call and a booking) — the kind of end-to-end flow worth diagramming rather than a vague "we use AI agents" claim.
Reel Script
Hook (16s / 38 words) An AI agent just called a doctor's office on the phone and booked you an appointment — using the same protocol Anthropic ships for connecting AI to tools. This isn't a chatbot demo. It's a live production integration shipping right now in healthcare.
Core Concept (60s / 138 words) The protocol here is MCP — Model Context Protocol — a standard way for an AI agent to call out to an external data source or tool without a developer hand-building a custom integration for every single system. Think of it like a USB port for AI agents: instead of wiring a new custom cable for every device, you plug into one standard socket. RevSpring, which runs a National Provider Directory, exposed that directory through an MCP server. PatientGenie's scheduling agents — the ones that already handle patient outreach over voice and text — now plug into that server directly. So instead of PatientGenie working off a stale, periodically-exported list of doctors, its agent queries RevSpring's directory live, pulling current phone numbers, specialties, and open availability at the moment it's trying to book someone.
Hands-On (70s / 160 words) Trace the actual flow. A health plan member needs an appointment. PatientGenie's agent queries RevSpring's MCP server and gets back live data: which providers match the need, their sub-specialty, their location, and their current availability — not a directory that was accurate three weeks ago. If that provider supports digital self-scheduling, the agent books it directly. If it doesn't — which is common for smaller practices — the agent doesn't just fail or hand off to a human. It places an outbound call, voice, to the provider's office, and books the appointment over the phone like a person would. That's agent-to-MCP-server-to-live-data-to-real-world-action, four steps, and the last one is a literal phone call completing a task a human scheduler used to do. The business loop closes too: RevSpring can now resell PatientGenie's agents to its own client base through that same connection, so both companies extend coverage from one integration.
Takeaway (24s / 56 words) This is what "agentic AI" is supposed to mean instead of the marketing version: a standard protocol, a live data source, and an agent that takes a real action — a phone call — to close the loop. If you're building agent integrations, MCP-to-live-directory-to-outbound-action is the pattern worth studying here.