Mercury
An operating system for a service business.
A full front office for service businesses, with an AI layer sitting on top that can actually act: an agent that calls tools, a live call copilot, and state machines that coordinate the work behind the scenes.
The live Mercury site.
Mercury is the ambitious one. It runs the whole front office (leads, scheduling, payments, payroll, loyalty, and configurable workflows) and then adds an AI layer that does more than answer questions. An orchestrator routes tool calls through a registry of integrations, a copilot rides along on live calls, and an event bus and state machines keep long-running processes on the rails.
It also ships a small server that exposes the platform to AI agents directly, so an assistant can book, look up, or update records through a defined interface instead of scraping a screen.
- A full front office: leads, CRM, booking, payments, payroll, loyalty, and configurable workflows.
- An AI layer with a tool-calling agent orchestrator, retrieval over your own data, and a live call copilot.
- A provider-adapter design with dozens of pluggable integrations and personal data encrypted at rest.
- A bundled server that exposes platform actions to AI agents through a defined interface.
- A React and TypeScript client over a Node and Express core.
The backend runs on Node and Express with MongoDB, Redis, and websockets. The AI layer is powered by both OpenAI and Anthropic models. Dozens of scheduled jobs drive the automation, and the orchestrator routes every tool call through the adapter registry so a new integration is a plug-in, not a rewrite.
- Node
- Express
- MongoDB
- Redis
- Socket.IO
- OpenAI
- Anthropic
- React
- TypeScript
Real figures from the build, not marketing math.