Caddy has a standard unix-like command line interface. Basic usage is:
caddy <command> [<args...>]
<carets> indicate parameters that get replaced by your input.
[brackets] indicate optional parameters.
... indicates a continuation, i.e. one or more parameters.
Adapts a config document to native JSON
Prints the environment
A simple but production-ready file server
Formats a Caddyfile
Hashes a password and outputs base64
View help for caddy commands
Lists the installed Caddy modules
Changes the config of the running Caddy process
A simple but production-ready reverse proxy
Starts the Caddy process in the foreground
Starts the Caddy process in the background
Stops the running Caddy process
Installs a certificate into local trust store(s)
Untrusts a certificate from local trust store(s)
Tests whether a config file is valid
Prints the version
caddy adapt [--config <path>] [--adapter <name>] [--pretty] [--validate]
Adapts a configuration to Caddy's native JSON config structure and writes the output to stdout, along with any warnings to stderr, then exits.
--config is the path to the config file. If omitted, assumes
Caddyfile in current directory if it exists; otherwise, this flag is required.
--adapter specifies the config adapter to use; default is
--pretty will format the output with indentation for human readability.
--validate will load and provision the adapted configuration to check for validity (but it will not actually start running the config).
Note that a config which is successfully adapted may still fail validation. For an example of this, use this Caddyfile:
localhost tls cert_notexist.pem key_notexist.pem
Try adapting it:
caddy adapt --config Caddyfile
It succeeds without error. Then try:
caddy adapt --config Caddyfile --validate adapt: validation: loading app modules: module name 'tls': provision tls: loading certificates: open cert_notexist.pem: no such file or directory
Even though that Caddyfile can be adapted to JSON without errors, the actual certificate and/or key files do not exist, so validation fails because that error arises during the provisioning phase. Thus, validation is a stronger error check than adaptation is.
To adapt a Caddyfile to JSON that you can easily read and tweak manually:
caddy adapt --config /path/to/Caddyfile --pretty
Prints the environment as seen by caddy, then exits. Can be useful when debugging init systems or process manager units like systemd.
caddy file-server [--root <path>] [--listen <addr>] [--domain <example.com>] [--browse] [--templates]
Spins up a simple but production-ready static file server.
--root specifies the root file path. Default is the current working directory.
--listen accepts a listener address. Default is
--domain is used, then
:443 will be the default.
--domain will only serve files through that hostname, and Caddy will attempt to serve it over HTTPS, so make sure any public DNS is configured properly first if it's a public domain name. The default port will be changed to 443.
--browse will enable directory listings if a directory without an index file is requested.
--templates will enable template rendering.
This command disables the admin API, making it easier to run multiple instances on a local development machine.
caddy fmt [<path>] [--overwrite]
Formats or prettifies a Caddyfile, then exits. The result is printed to stdout unless
--overwrite is used.
<path> specifies the path to the Caddyfile. If omitted, a file named Caddyfile in the current directory is assumed instead.
--overwrite causes the result to be written to the input file instead of being printed to the terminal.
caddy hash-password --plaintext <password> [--algorithm <name>] [--salt <string>]
Hashes a password and writes the output to stdout in base64 encoding, then exits.
--plaintext is the plaintext form of the password.
--algorithm may be bcrypt or scrypt. Default is bcrypt.
--salt is used only if the algorithm requires an external salt (like scrypt).
caddy help [<command>]
Prints CLI help text, optionally for a specific subcommand, then exits.
caddy list-modules [--versions]
Prints the Caddy modules that are installed, optionally with version information from their associated Go modules, then exits.
NOTE: Due to a bug in Go, version information is only available if Caddy is built as a dependency and not as the main module. TODO: Link to docs that explain how to build Caddy with version info
caddy reload [--config <path>] [--adapter <name>] [--address <interface>]
Gives the running Caddy instance a new configuration. This has the same effect as POSTing a document to the /load endpoint, but this command is convenient for simple workflows revolving around config files. Compared to the
run commands, this single command is the correct, semantic way to change/reload the running configuration.
Because this command uses the API, the admin endpoint must not be disabled.
--config is the config file to apply. If not specified, it will try a file called
Caddyfile in the current working directory and, if it exists, it will adapt it using the
caddyfile config adapter; otherwise, it is an error if there is no config file to load.
--adapter specifies a config adapter to use, if any.
--address needs to be used if the admin endpoint is not listening on the default address and if it is different from the address in the provided config file. Note that only TCP addresses are supported at this time.
caddy reverse-proxy [--from <addr>] --to <addr> [--change-host-header]
Spins up a simple but production-ready reverse proxy.
--from is the address to proxy from.
--to is the address to proxy to.
--change-host-header will cause Caddy to change the Host header from the incoming value to the address of the upstream.
--to parameters can be URLs, as scheme and domain name will be inferred from the provided URL (paths and query strings ignored). Or they can be a simple network address and not a complete URL.
This command disables the admin API so it is easier to run multiple instances on a local development machine.
caddy run [--config <path>] [--adapter <name>] [--environ] [--resume] [--watch]
Runs Caddy and blocks indefinitely; i.e. "daemon" mode.
--config specifies an initial config file to immediately load and use. If no config is specified, Caddy will run with a blank configuration and use default settings for the admin API endpoints, which can be used to feed it new configuration. As a special case, if the current working directory has a file called "Caddyfile" and the
caddyfile config adapter is plugged in (default), then that file will be loaded and used to configure Caddy, even without any command line flags.
--adapter is the name of the config adapter to use when loading the initial config, if any. This flag is not necessary if the
--config filename starts with "Caddyfile" which assumes the
caddyfile adapter. Otherwise, this flag is required if the provided config file is not in Caddy's native JSON format. Any warnings will be printed to the log, but beware that any adaptation without errors will immediately be used, even if there are warnings. If you want to review the results of the adaptation first, use the
caddy adapt subcommand.
--environ prints out the environment before starting. This is the same as the
caddy environ command, but does not exit after printing.
--resume uses the last loaded configuration, overriding the
--config flag (if present) if a previous config was saved. Using this flag guarantees config durability through machine reboots or process restarts. It is most useful in API-heavy deployments.
--watch will watch the config file and automatically reload it after it changes. ⚠️ This feature is intended for use only in local development environments!
caddy start [--config <path>] [--adapter <name>] [--watch]
caddy run, but in the background. This command only blocks until the background process is running successfully (or fails to run), then returns.
Use of this command is discouraged with system services or on Windows. On Windows, the child process will remain attached to the terminal, so closing the window will forcefully stop Caddy, which is not obvious.
caddy stop [--address <interface>]
Gracefully stops the running Caddy process (other than the process of the stop command) and causes it to exit. It uses the /stop endpoint of the admin API to perform a graceful shutdown.
--address can be used if the running instance's admin API is not on the default port; an alternate address can be specified here.
Installs the root certificate for Caddy's default internal CA (named "local") into the local trust store(s); intended for development environments only. May prompt for a password if there are not already sufficient privileges.
This command is often unnecessary. Because Caddy will install its root certificate into local trust stores automatically when first needed, this command is only useful if you need to pre-install the certificates while you have elevated privileges, like during system provisioning in automated environments.
caddy untrust [--ca <id>] [--cert <path>]
Untrusts a root certificate from the local trust store(s). Intended for development environments only. Specify either the
--cert flags, but not both. If neither are specified, Caddy's default CA (
--ca specifies the ID of the Caddy CA to untrust. The default CA's ID is
--cert specifies the path to the PEM-encoded certificate file to uninstall.
caddy validate [--config <path>] [--adapter <name>]
Validates a configuration file, then exits. This command deserializes the config, then loads and provisions all of its modules as if to start the config, but the config is not actually started. This exposes errors in a configuration that arise during loading or provisioning phases and is a stronger error check than merely serializing a config as JSON.
--config is the config file to validate. Default is the
Caddyfile in the current directory, if any.
--adapter is the name of the config adapter to use, if the config file is not in Caddy's native JSON format. If the config file starts with
caddyfile adapter is used by default.
Prints the version and exits.