Documentation
a project

Request Matchers

Request matchers can be used to filter (or classify) requests by various criteria.

Syntax

In the Caddyfile, a matcher token immediately following the directive can limit that directive's scope. The matcher token can be one of these forms:

  1. * to match all requests (wildcard; default).
  2. /path start with a forward slash to match a request path.
  3. @name to specify a named matcher.

If a directive supports matchers, it will appear as [<matcher>] in its syntax documentation. Matcher tokens are usually optional, denoted by [ ]. If the matcher token is omitted, it is the same as a wildcard matcher (*).

Examples

This directive applies to all HTTP requests:

reverse_proxy localhost:9000

And this is the same (* is unnecessary here):

reverse_proxy * localhost:9000

But this directive applies only to requests having a path starting with /api/:

reverse_proxy /api/* localhost:9000

To match on anything other than a path, define a named matcher and refer to it using @name:

@postfoo {
	method POST
	path /foo/*
}
reverse_proxy @postfoo localhost:9000

Wildcard matchers

The wildcard (or "catch-all") matcher * matches all requests, and is only needed if a matcher token is required. For example, if the first argument you want to give a directive also happens to be a path, it would look exactly like a path matcher! So you can use a wildcard matcher to disambiguate, for example:

root * /home/www/mysite

Otherwise, this matcher is not often used. We generally recommend omitting it if syntax doesn't require it.

Path matchers

Matching by URI path is the most common way to match requests, so the matcher can be inlined, like this:

redir /old.html /new.html

Path matcher tokens must start with a forward slash /.

Path matching is an exact match by default, not a prefix match. You must append a * for a fast prefix match. Note that /foo* will match /foo and /foo/ as well as /foobar; you might actually want /foo/* instead.

Named matchers

All matchers that are not path or wildcard matchers must be named matchers. This is a matcher that is defined outside of any particular directive, and can be reused.

Defining a matcher with a unique name gives you more flexibility, allowing you to combine any available matchers into a set:

@name {
	...
}

or, if there is only one matcher in the set, you can put it on the same line:

@name ...

Then you can use the matcher like so, by specifying it as the first argument to a directive:

directive @name

For example, this proxies websocket requests to localhost:6001, and other requests to localhost:8080. It matches requests that have a header field named Connection containing Upgrade, and another field named Upgrade with exactly websocket:

example.com {
	@websockets {
		header Connection *Upgrade*
		header Upgrade    websocket
	}
	reverse_proxy @websockets localhost:6001

	reverse_proxy localhost:8080
}

If the matcher set consists of only one matcher, a one-liner syntax also works:

@post method POST
reverse_proxy @post localhost:6001

As a special case, the expression matcher may be used without specifying its name as long as one quoted argument (the CEL expression itself) follows the matcher name:

@not-found `{err.status_code} == 404`

Like directives, named matcher definitions must go inside the site blocks that use them.

A named matcher definition constitutes a matcher set. Matchers in a set are AND'ed together; i.e. all must match. For example, if you have both a header and path matcher in the set, both must match.

Multiple matchers of the same type may be merged (e.g. multiple path matchers in the same set) using boolean algebra (AND/OR), as described in their respective sections below.

For more complex boolean matching logic, it's recommended to the expression matcher to write a CEL expression, which supports and &&, or ||, and parentheses ( ).

Standard matchers

Full matcher documentation can be found in each respective matcher module's docs.

Requests can be matched the following ways:

client_ip

client_ip <ranges...>

expression client_ip('<ranges...>')

By the client IP address. Accepts exact IPs or CIDR ranges. IPv6 zones are supported.

This matcher is best used when the trusted_proxies global option is configured, otherwise it acts identically to the remote_ip matcher. Only requests from trusted proxies will have their client IP parsed at the start of the request; untrusted requests will use the remote IP address of the immediate peer.

As a shortcut, private_ranges can be used to match all private IPv4 and IPv6 ranges. It's the same as specifying all of these ranges: 192.168.0.0/16 172.16.0.0/12 10.0.0.0/8 127.0.0.1/8 fd00::/8 ::1

There can be multiple client_ip matchers per named matcher, and their ranges will be merged and OR'ed together.

Example:

Match requests from private IPv4 addresses:

@private-ipv4 client_ip 192.168.0.0/16 172.16.0.0/12 10.0.0.0/8 127.0.0.1/8

This matcher is commonly paired with the not matcher to invert the match. For example, to abort all connections from public IPv4 and IPv6 addresses (which is the inverse of all private ranges):

example.com {
	@denied not client_ip private_ranges
	abort @denied

	respond "Hello, you must be from a private network!"
}

In a CEL expression, it would look like this:

@my-friends `client_ip('12.23.34.45', '23.34.45.56')`

expression

expression <cel...>

By any CEL (Common Expression Language) expression that returns true or false.

Caddy placeholders (or Caddyfile shorthands) may be used in these CEL expressions, as they are preprocessed and converted to regular CEL function calls before being interpreted by the CEL environment.

Most other request matchers can also be used in expressions as functions, which allows for more flexibility for boolean logic than outside expressions. See the documentation for each matcher for the supported syntax within CEL expressions.

For convenience, the matcher name may be omitted if defining a named matcher that consists solely of a CEL expression. The CEL expression must be quoted (backticks or heredocs recommended). This reads quite nicely:

@mutable `{method}.startsWith("P")`

In this case the CEL matcher is assumed.

Examples:

Match requests whose methods start with P, e.g. PUT or POST:

@methods expression {method}.startsWith("P")

Match requests where handler returned error status code 404, would be used in conjunction with the handle_errors directive:

@404 expression {err.status_code} == 404

Match requests where the path matches one of two different regular expressions; this is only possible to write using an expression, because the path_regexp matcher can normally only exist once per named matcher:

@user expression path_regexp('^/user/(\w*)') || path_regexp('^/(\w*)')

Or the same, omitting the matcher name, and wrapping in backticks so it's parsed as a single token:

@user `path_regexp('^/user/(\w*)') || path_regexp('^/(\w*)')`

You may use heredoc syntax to write multi-line CEL expressions:

@api <<CEL
	{method} == "GET"
	&& {path}.startsWith("/api/")
	CEL
respond @api "Hello, API!"

file

file {
	root       <path>
	try_files  <files...>
	try_policy first_exist|smallest_size|largest_size|most_recently_modified
	split_path <delims...>
}
file <files...>

expression `file({
	'root': '<path>',
	'try_files': ['<files...>'],
	'try_policy': 'first_exist|smallest_size|largest_size|most_recently_modified',
	'split_path': ['<delims...>']
})`
expression file('<files...>')

By files.

  • root defines the directory in which to look for files. Default is the current working directory, or the root variable ({http.vars.root}) if set (can be set via the root directive).

  • try_files checks files in its list that match the try_policy.

    To match directories, append a trailing forward slash / to the path. All file paths are relative to the site root, and glob patterns will be expanded.

    If the try_policy is first_exist (the default), then the last item in the list may be a number prefixed by = (e.g. =404), which as a fallback, will emit an error with that code; the error can be caught and handled with handle_errors.

  • try_policy specifies how to choose a file. Default is first_exist.

    • first_exist checks for file existence. The first file that exists is selected.

    • smallest_size chooses the file with the smallest size.

    • largest_size chooses the file with the largest size.

    • most_recently_modified chooses the file that was most recently modified.

  • split_path will cause the path to be split at the first delimiter in the list that is found in each filepath to try. For each split value, the left-hand side of the split including the delimiter itself will be the filepath that is tried. For example, /remote.php/dav/ using a delimiter of .php would try the file /remote.php. Each delimiter must appear at the end of a URI path component in order to be used as a split delimiter. This is a niche setting and is mostly used when serving PHP sites.

Because try_files with a policy of first_exist is so common, there is a one-line shortcut for that:

file <files...>

An empty file matcher (one with no files listed after it) will see if the requested file—verbatim from the URI, relative to the site root—exists. This is effectively the same as file {path}.

Upon matching, four new placeholders will be made available:

  • {file_match.relative} The root-relative path of the file. This is often useful when rewriting requests.
  • {file_match.absolute} The absolute path of the matched file, including the root.
  • {file_match.type} The type of file, file or directory.
  • {file_match.remainder} The portion remaining after splitting the file path (if split_path is configured)

Examples:

Match requests where the path is a file that exists:

@file file

Match requests where the path followed by .html is a file that exists, or if not, where the path is a file that exists:

@html file {
	try_files {path}.html {path} 
}

Same as above, except using the one-line shortcut, and falling back to emitting a 404 error if a file is not found:

@html-or-error file {path}.html {path} =404

Some more examples using CEL expressions. Keep in mind that placeholders are preprocessed and converted to regular CEL function calls before being interpreted by the CEL environment, so concatenation is used here. Additionally, the long-form must be used if concatenating with placeholders due to current parsing limitations:

@file `file()`
@first `file({'try_files': [{path}, {path} + '/', 'index.html']})`
@smallest `file({'try_policy': 'smallest_size', 'try_files': ['a.txt', 'b.txt']})`

header <field> [<value> ...]

expression header({'<field>': '<value>'})

By request header fields.

  • <field> is the name of the HTTP header field to check.
    • If prefixed with !, the field must not exist to match (omit value arg).
  • <value> is the value the field must have to match. One or more may be specified.
    • If prefixed with *, it performs a fast suffix match (appears at the end).
    • If suffixed with *, it performs a fast prefix match (appears at the start).
    • If enclosed by *, it performs a fast substring match (appears anywhere).
    • Otherwise, it is a fast exact match.

Different header fields within the same set are AND-ed. Multiple values per field are OR'ed.

Note that header fields may be repeated and have different values. Backend applications MUST consider that header field values are arrays, not singular values, and Caddy does not interpret meaning in such quandaries.

Example:

Match requests with the Connection header containing Upgrade:

@upgrade header Connection *Upgrade*

Match requests with the Foo header containing bar OR baz:

@foo {
	header Foo bar
	header Foo baz
}

Match requests that do not have the Foo header field at all:

@not_foo header !Foo

Using an CEL expression, match WebSocket requests by checking for the Connection header containing Upgrade and the Upgrade header equalling websocket:

@websockets `header({'Connection':'*Upgrade*','Upgrade':'websocket'})`

header_regexp

header_regexp [<name>] <field> <regexp>

expression header_regexp('<name>', '<field>', '<regexp>')
expression header_regexp('<field>', '<regexp>')

Like header, but supports regular expressions.

The regular expression language used is RE2, included in Go. See the RE2 syntax reference and the Go regexp syntax overview.

As of v2.8.0, if name is not provided, the name will be taken from the named matcher's name. For example a named matcher @foo will cause this matcher to be named foo. The main advantage of specifying a name is if more than one regexp matcher (e.g. header_regexp and path_regexp, or multiple different header fields) is used in the same named matcher.

Capture groups can be accessed via placeholder in directives after matching:

  • {re.<name>.<capture_group>} where:

    • <name> is the name of the regular expression,
    • <capture_group> is either the name or number of the capture group in the expression.

  • {re.<capture_group>} without a name, is also populated for convenience. The caveat is that if multiple regexp matchers are used in sequence, then the placeholder values will be overwritten by the next matcher.

Capture group 0 is the full regexp match, 1 is the first capture group, 2 is the second capture group, and so on. So {re.foo.1} or {re.1} will both hold the value of the first capture group.

Only one regular expression is supported per header field, since regexp patterns cannot be merged; if you need more, consider using an expression matcher. Matches against multiple different header fields will be AND'ed.

Example:

Match requests where the Cookie header contains login_ followed by a hex string, with a capture group that can be accessed with {re.login.1} or {re.1}.

@login header_regexp login Cookie login_([a-f0-9]+)

This can be simplified by omitting the name, which will be inferred from the named matcher:

@login header_regexp Cookie login_([a-f0-9]+)

Or the same, using a CEL expression:

@login `header_regexp('login', 'Cookie', 'login_([a-f0-9]+)')`

host

host <hosts...>

expression host('<hosts...>')

Matches request by the Host header field of the request.

Since most site blocks already indicate hosts in the address of the site, this matcher is more commonly used in site blocks that use a wildcard hostname (see the wildcard certificates pattern), but where hostname-specific logic is required.

Multiple host matchers will be OR'ed together.

Example:

Matching one subdomain:

@sub host sub.example.com

Matching the apex domain and a subdomain:

@site host example.com www.example.com

Multiple subdomains using a CEL expression:

@app `host('app1.example.com', 'app2.example.com')`

method

method <verbs...>

expression method('<verbs...>')

By the method (verb) of the HTTP request. Verbs should be uppercase, like POST. Can match one or many methods.

Multiple method matchers will be OR'ed together.

Examples:

Match requests with the GET method:

@get method GET

Match requests with the PUT or DELETE methods:

@put-delete method PUT DELETE

Match read-only methods using a CEL expression:

@read `method('GET', 'HEAD', 'OPTIONS')`

not

not <matcher>

or, to negate multiple matchers which get AND'ed, open a block:

not {
	<matchers...>
}

The results of the enclosed matchers will be negated.

Examples:

Match requests with paths that do NOT begin with /css/ OR /js/.

@not-assets {
	not path /css/* /js/*
}

Match requests WITH NEITHER:

  • an /api/ path prefix, NOR
  • the POST request method

i.e. must have none of these to match:

@with-neither {
	not path /api/*
	not method POST
}

Match requests WITHOUT BOTH:

  • an /api/ path prefix, AND
  • the POST request method

i.e. must have neither or either of these to match:

@without-both {
	not {
		path /api/*
		method POST
	}
}

There's no CEL expression for this matcher, because you may use the ! operator for negation instead. For example:

@without-both `!path('/api*') && !method('POST')`

Which is the same as this, using parentheses:

@without-both `!(path('/api*') || method('POST'))`

path

path <paths...>

expression path('<paths...>')

By request path (the path component of the request URI). Path matches are exact but case-insensitive. Wildcards * may be used:

  • At the end only, for a prefix match (/prefix/*)
  • At the beginning only, for a suffix match (*.suffix)
  • On both sides only, for a substring match (*/contains/*)
  • In the middle only, for a globular match (/accounts/*/info)

Slashes are significant. For example, /foo* will match /foo, /foobar, /foo/, and /foo/bar, but /foo/* will not match /foo or /foobar.

Request paths are cleaned to resolve directory traversal dots before matching. Additionally, multiple slashes are merged unless the match pattern has multiple slashes. In other words, /foo will match /foo and //foo, but //foo will only match //foo.

Because there are multiple escaped forms of any given URI, the request path is normalized (URL-decoded, unescaped) except for those escape sequences at positions where escape sequences are also present in the match pattern. For example, /foo/bar matches both /foo/bar and /foo%2Fbar, but /foo%2Fbar will match only /foo%2Fbar, because the escape sequence is explicitly given in the configuration.

The special wildcard escape %* can also be used instead of * to leave its matching span escaped. For example, /bands/*/* will not match /bands/AC%2FDC/T.N.T because the path will be compared in normalized space where it looks like /bands/AC/DC/T.N.T, which does not match the pattern; however, /bands/%*/* will match /bands/AC%2FDC/T.N.T because the span represented by %* will be compared without decoding escape sequences.

Multiple paths will be OR'ed together.

Examples:

Match multiple directories and their contents:

@assets path /js/* /css/* /images/*

Match a specific file:

@favicon path /favicon.ico

Match file extensions:

@extensions path *.js *.css

With a CEL expression:

@assets `path('/js/*', '/css/*', '/images/*')`

path_regexp

path_regexp [<name>] <regexp>

expression path_regexp('<name>', '<regexp>')
expression path_regexp('<regexp>')

Like path, but supports regular expressions. Runs against the URI-decoded/unescaped path.

The regular expression language used is RE2, included in Go. See the RE2 syntax reference and the Go regexp syntax overview.

As of v2.8.0, if name is not provided, the name will be taken from the named matcher's name. For example a named matcher @foo will cause this matcher to be named foo. The main advantage of specifying a name is if more than one regexp matcher (e.g. path_regexp and header_regexp) is used in the same named matcher.

Capture groups can be accessed via placeholder in directives after matching:

  • {re.<name>.<capture_group>} where:

    • <name> is the name of the regular expression,
    • <capture_group> is either the name or number of the capture group in the expression.

  • {re.<capture_group>} without a name, is also populated for convenience. The caveat is that if multiple regexp matchers are used in sequence, then the placeholder values will be overwritten by the next matcher.

Capture group 0 is the full regexp match, 1 is the first capture group, 2 is the second capture group, and so on. So {re.foo.1} or {re.1} will both hold the value of the first capture group.

There can only be one path_regexp pattern per named matcher, since this matcher cannot be merged with itself; if you need more, consider using an expression matcher.

Example:

Match requests where the path ends a 6 character hex string followed by .css or .js as the file extension, with capture groups (parts enclosed in ( )), that can be accessed with {re.static.1} and {re.static.2} (or {re.1} and {re.2}), respectively:

@static path_regexp static \.([a-f0-9]{6})\.(css|js)$

This can be simplified by omitting the name, which will be inferred from the named matcher:

@static path_regexp \.([a-f0-9]{6})\.(css|js)$

Or the same, using a CEL expression, also validating that the file exists on disk:

@static `path_regexp('\.([a-f0-9]{6})\.(css|js)$') && file()`

protocol

protocol http|https|grpc|http/<version>[+]

expression protocol('http|https|grpc|http/<version>[+]')

By request protocol. A broad protocol name such as http, https, or grpc can be used; or specific or minimum HTTP versions such as http/1.1 or http/2+.

There can only be one protocol matcher per named matcher.

Example:

Match requests using HTTP/2:

@http2 protocol http/2+

With a CEL expression:

@http2 `protocol('http/2+')`

query

query <key>=<val>...

expression query({'<key>': '<val>'})
expression query({'<key>': ['<vals...>']})

By query string parameters. Should be a sequence of key=value pairs. Keys are matched exactly (case-sensitively) but also support * to match any value. Values can use placeholders.

There can be multiple query matchers per named matcher, and pairs with the same keys will be OR'ed together. Different keys will be AND'ed together. So, all keys in the matcher must have at least one matching value.

Illegal query strings (bad syntax, unescaped semicolons, etc.) will fail to parse and thus will not match.

NOTE: Query string parameters are arrays, not singular values. This is because repeated keys are valid in query strings, and each one may have a different value. This matcher will match for a key if any one of its configured values is assigned in the query string. Backend applications using query strings MUST take into consideration that query string values are arrays and can have multiple values.

Example:

Match a q query parameter with any value:

@search query q=*

Match a sort query parameter with the value asc or desc:

@sorted query sort=asc sort=desc

Matching both q and sort, with a CEL expression:

@search-sort `query({'sort': ['asc', 'desc'], 'q': '*'})`

remote_ip

remote_ip <ranges...>

expression remote_ip('<ranges...>')

By remote IP address (i.e. the IP address of the immediate peer). Accepts exact IPs or CIDR ranges. IPv6 zones are supported.

As a shortcut, private_ranges can be used to match all private IPv4 and IPv6 ranges. It's the same as specifying all of these ranges: 192.168.0.0/16 172.16.0.0/12 10.0.0.0/8 127.0.0.1/8 fd00::/8 ::1

if you wish to match the "real IP" of the client, as parsed from HTTP headers, use the client_ip matcher instead.

There can be multiple remote_ip matchers per named matcher, and their ranges will be merged and OR'ed together.

Example:

Match requests from private IPv4 addresses:

@private-ipv4 remote_ip 192.168.0.0/16 172.16.0.0/12 10.0.0.0/8 127.0.0.1/8

This matcher is commonly paired with the not matcher to invert the match. For example, to abort all connections from public IPv4 and IPv6 addresses (which is the inverse of all private ranges):

example.com {
	@denied not remote_ip private_ranges
	abort @denied

	respond "Hello, you must be from a private network!"
}

In a CEL expression, it would look like this:

@my-friends `remote_ip('12.23.34.45', '23.34.45.56')`

vars

vars <variable> <values...>

By the value of a variable in the request context, or the value of a placeholder. Multiple values may be specified to match any of those possible values (OR'ed).

The <variable> argument may be either a variable name or a placeholder in curly braces { }. (Placeholders are not expanded in the first parameter.)

This matcher is most useful when paired with the map directive which sets outputs, or with plugins which set some information in the request context.

Example:

Match an output of the map directive named magic_number for the values 3 or 5:

vars {magic_number} 3 5

Match an arbitrary placeholder's value, i.e. the authenticated user's ID, either Bob or Alice:

vars {http.auth.user.id} Bob Alice

vars_regexp

vars_regexp [<name>] <variable> <regexp>

Like vars, but supports regular expressions.

The regular expression language used is RE2, included in Go. See the RE2 syntax reference and the Go regexp syntax overview.

As of v2.8.0, if name is not provided, the name will be taken from the named matcher's name. For example a named matcher @foo will cause this matcher to be named foo. The main advantage of specifying a name is if more than one regexp matcher (e.g. vars_regexp and header_regexp) is used in the same named matcher.

Capture groups can be accessed via placeholder in directives after matching:

  • {re.<name>.<capture_group>} where:

    • <name> is the name of the regular expression,
    • <capture_group> is either the name or number of the capture group in the expression.

  • {re.<capture_group>} without a name, is also populated for convenience. The caveat is that if multiple regexp matchers are used in sequence, then the placeholder values will be overwritten by the next matcher.

Capture group 0 is the full regexp match, 1 is the first capture group, 2 is the second capture group, and so on. So {re.foo.1} or {re.1} will both hold the value of the first capture group.

Only one regular expression is supported per variable name, since regexp patterns cannot be merged; if you need more, consider using an expression matcher. Matches against multiple different variables will be AND'ed.

Example:

Match an output of the map directive named magic_number for a value starting with 4, capturing the value in a capture group that can be accessed with {re.magic.1} or {re.1}:

@magic vars_regexp magic {magic_number} ^(4.*)

This can be simplified by omitting the name, which will be inferred from the named matcher:

@magic vars_regexp {magic_number} ^(4.*)