DiscoverVulcain | Server Push Web APIs


Install the hub

Managed and HA Versions

Managed and High Availability versions of are available, give them a try!

Prebuilt Binary

The hub is available as a custom build of the Caddy web server including the module.

First, download the archive corresponding to your operating system and architecture from the release page, extract the archive and open a shell in the resulting directory.

Note: Mac OS users must use the Darwin binary.

Then, to start the Hub in development mode on Linux and Mac OS X, run:

./mercure run -config

On Windows, start PowerShell, go into the extracted directory and run:

$env:MERCURE_PUBLISHER_JWT_KEY='!ChangeMe!'; $env:MERCURE_SUBSCRIBER_JWT_KEY='!ChangeMe!'; .\mercure.exe run -config

Note: The Windows Defender Firewall will ask you if you want to allow mercure.exe to communicate through it. Allow it for both public and private networks. If you use an antivirus, or another firewall software, be sure to whitelist mercure.exe.

The server is now available on https://localhost (TLS is automatically enabled, learn how to disable it). In development mode, anonymous subscribers are allowed and the debug UI is available on https://localhost/.well-known/mercure/ui/.

Note: if you get an error similar to bind: address already in use, it means that the port 80 or 443 is already used by another service (the usual suspects are Apache and NGINX). Before starting Mercure, stop the service using the port(s) first, or set the SERVER_NAME environment variable to use a free port (ex: SERVER_NAME=:3000).

To run the server in production mode, run this command:

./mercure run

In production mode, the debugger UI is disabled and anonymous subscribers aren't allowed. To change these default settings, learn how to configure the hub.

When the server is up and running, the following endpoints are available:

  • POST https://localhost/.well-known/mercure: to publish updates
  • GET https://localhost/.well-known/mercure: to subscribe to updates

See the protocol for more details about these endpoints.

To compile the development version, see

Docker Image

A Docker image is available on Docker Hub. The following command is enough to get a working server in development mode:

docker run \
    -p 80:80 \
    -p 443:443 \
    dunglas/mercure caddy run -config /etc/caddy/

The server is then available on https://localhost. Anonymous subscribers are allowed and the debugger UI is available on https://localhost/.well-known/mercure/ui/.

In production, simply run:

docker run \
    -p 80:80 \
    -p 443:443 \

HTTPS support is automatically enabled. If you run the Mercure hub behind a reverse proxy such as NGINX, you usually want to use unencrypted HTTP. This can be done like that:

docker run \
    -e SERVER_NAME=':80' \
    -p 80:80 \

The Docker image is based on the Caddy Server Docker image. See the configuration section and the documentation of the Docker image for Caddy to learn how to configure it to fit your needs.

Docker Compose

If you prefer to use docker-compose to run the hub, here's a sample service definition:

# docker-compose.yml
version: "3.7"

    image: dunglas/mercure
    restart: unless-stopped
      # Uncomment the following line to disable HTTPS
      #SERVER_NAME: ':80'
    # Uncomment the following line to enable the development mode
    #command: /usr/bin/caddy run -config /etc/caddy/
      - "80:80"
      - "443:443"
      - caddy_data:/data
      - caddy_config:/config


Arch Linux is available on the AUR, you can install it with your favorite AUR wrapper:

yay -S mercure

Or download the PKGBUILD and compile and install it: makepkg -sri.

Custom Caddy Build

It's also possible to download Caddy with Mercure and other modules included, or to build your own binaries using xcaddy:

xcaddy build --with

The is shipped by several popular services and frameworks, including Symfony and API Platform. Refer to their documentations to get started.