llms.txt Content
# Mercure.rocks
> Mercure streams real-time updates from your servers to any client over Server-Sent Events. Fast, reliable, battery-efficient, and a modern replacement for WebSockets and the higher-level libraries built on top of them. Free and open source, with managed Cloud and Enterprise editions.
Mercure shines for any architecture that needs to push server-driven updates to clients in real time: live dashboards, collaborative tools, notifications, presence, and especially AI applications. ChatGPT, Claude, and the OpenAI and Anthropic APIs all stream responses over Server-Sent Events; Mercure speaks the same transport, so it is a natural broadcast layer in front of LLMs and AI agents (token streaming, intermediate reasoning steps, tool-call updates, and any event the agent emits).
It is equally useful for adding streaming and asynchronous capabilities to REST and GraphQL APIs. Because Mercure runs on top of plain HTTP, it is natively supported by modern web browsers, mobile applications and IoT devices. The protocol supports JWT-based authorization for private updates, automatic reconnection with replay of missed messages, presence detection, an event store, end-to-end encryption, and serverless-friendly publishing via a simple POST request.
Three ways to run Mercure: the open source reference Hub (self-hosted), **Mercure Cloud** (fully managed, with Hobby, Pro, Pro+ and Pro++ tiers), and **Mercure Enterprise** (self-hosted on the customer's own infrastructure). Mercure is backed by an open IETF-style specification anyone can implement.
## Documentation
- [Getting Started](https://mercure.rocks/docs/getting-started): install and run a Mercure Hub, publish and subscribe to updates
- [Hub installation](https://mercure.rocks/docs/hub/install): install the reference Hub on a server, container or as a Caddy module
- [Hub configuration](https://mercure.rocks/docs/hub/config): runtime configuration, JWT keys, transports, CORS
- [Mercure Cloud](https://mercure.ro