HTTPie (pronounced aitch-tee-tee-pie) is a command line HTTP client. Its goal is to make CLI interaction with web services as human-friendly as possible. It provides a simple http command that allows for sending arbitrary HTTP requests using a simple and natural syntax, and displays colorized output. HTTPie can be used for testing, debugging, and generally interacting with HTTP servers.
- Expressive and intuitive syntax
- Formatted and colorized terminal output
- Built-in JSON support
- Forms and file uploads
- HTTPS, proxies, and authentication
- Arbitrary request data
- Custom headers
- Persistent sessions
- Wget-like downloads
- Python 2.6, 2.7 and 3.x support
- Linux, Mac OS X and Windows support
- Plugins
- Documentation
- Test coverage
On macOS, HTTPie can be installed via Homebrew (recommended):
A MacPorts port is also available:
Most Linux distributions provide a package that can be installed using the system package manager, for example:
A universal installation method (that works on Windows, Mac OS X, Linux, …, and always provides the latest version) is to use pip:
(If pip installation fails for some reason, you can try easy_install httpie as a fallback.)
Although Python 2.6 and 2.7 are supported as well, it is recommended to install HTTPie against the latest Python 3.x whenever possible. That will ensure that some of the newer HTTP features, such as SNI (Server Name Indication), work out of the box. Python 3 is the default for Homebrew installations starting with version 0.9.4. To see which version HTTPie uses, run http --debug.
You can also instead of the latest the latest unreleased development version directly from the master branch on GitHub. It is a work-in-progress of a future stable release so the experience might be not as smooth.
On macOS you can install it with Homebrew:
Otherwise with pip:
Verify that now we have the current development version identifier with the -dev suffix, for example:
Hello World:
Synopsis:
See also http --help.
Custom HTTP method, HTTP headers and JSON data:
Submitting forms:
See the request that is being sent using one of the output options:
Use Github API to post a comment on an issue with authentication:
Upload a file using redirected input:
Download a file and save it via redirected output:
Download a file wget style:
Use named sessions to make certain aspects or the communication persistent between requests to the same host:
Set a custom Host header to work around missing DNS records:
The name of the HTTP method comes right before the URL argument:
Which looks similar to the actual Request-Line that is sent:
DELETE /todos/7 HTTP/1.1When the METHOD argument is omitted from the command, HTTPie defaults to either GET (with no request data) or POST (with request data).
The only information HTTPie needs to perform a request is a URL. The default scheme is, somewhat unsurprisingly, http://, and can be omitted from the argument – http example.org works just fine.
If you find yourself manually constructing URLs with querystring parameters on the terminal, you may appreciate the param==value syntax for appending URL parameters. With that, you don't have to worry about escaping the & separators for your shell. Also, special characters in parameter values, will also automatically escaped (HTTPie otherwise expects the URL to be already escaped). To search for HTTPie logo on Google Images you could use this command:
GET /?search=HTTPie+logo&tbm=isch HTTP/1.1Additionally, curl-like shorthand for localhost is supported. This means that, for example :3000 would expand to http://localhost:3000 If the port is omitted, then port 80 is assumed.
GET /foo HTTP/1.1
Host: localhostGET /bar HTTP/1.1
Host: localhost:3000GET / HTTP/1.1
Host: localhostYou can use the --default-scheme <URL_SCHEME> option to create shortcuts for other protocols than HTTP:
There are a few different request item types that provide a convenient mechanism for specifying HTTP headers, simple JSON and form data, files, and URL parameters.
They are key/value pairs specified after the URL. All have in common that they become part of the actual request that is sent and that their type is distinguished only by the separator used: :, =, :=, ==, @, =@, and :=@. The ones with an @ expect a file path as value.
| Item Type | Description |
|---|---|
HTTP Headers Name:Value |
Arbitrary HTTP header, e.g. X-API-Token:123. |
URL parameters name==value |
Appends the given name/value pair as a query string parameter to the URL. The == separator is used. |
Data Fields field=value, field=@file.txt |
Request data fields to be serialized as a JSON object (default), or to be form-encoded (--form, -f). |
Raw JSON fields field:=json, field:=@file.json |
Useful when sending JSON and one or more fields need to be a Boolean, Number, nested Object, or an Array, e.g., meals:='["ham","spam"]' or pies:=[1,2,3] (note the quotes). |
Form File Fields field@/dir/file |
Only available with --form, -f. For example screenshot@~/Pictures/img.png. The presence of a file field results in a multipart/form-data request. |
Note that data fields aren't the only way to specify request data: Redirected input is a mechanism for passing arbitrary data request request.
You can use \ to escape characters that shouldn't be used as separators (or parts thereof). For instance, foo\==bar will become a data key/value pair (foo= and bar) instead of a URL parameter.
Often it is necessary to quote the values, e.g. foo='bar baz'.
If any of the field names or headers starts with a minus (e.g., -fieldname), you need to place all such items after the special token -- to prevent confusion with --arguments:
POST /post HTTP/1.1
-Unusual-Header: bar
Content-Type: application/json
{
"-name-starting-with-dash": "value"
}JSON is the lingua franca of modern web services and it is also the implicit content type HTTPie by default uses.
Simple example:
PUT / HTTP/1.1
Accept: application/json, */*
Accept-Encoding: gzip, deflate
Content-Type: application/json
Host: example.org
{
"name": "John",
"email": "john@example.org"
}If your command includes some data request items, they are serialized as a JSON object by default. HTTPie also automatically sets the following headers, both of which can be overwritten:
Content-Type |
application/json |
Accept |
application/json, */* |
You can use --json, -j to explicitly set Accept to application/json regardless of whether you are sending data (it's a shortcut for setting the header via the usual header notation: http url Accept:'application/json, */*'). Additionally, HTTPie will try to detect JSON responses even when the Content-Type is incorrectly text/plain or unknown.
Non-string fields use the := separator, which allows you to embed raw JSON into the resulting object. Text and raw JSON files can also be embedded into fields using =@ and :=@:
PUT /person/1 HTTP/1.1
Accept: application/json, */*
Content-Type: application/json
Host: api.example.com
{
"age": 29,
"hobbies": [
"http",
"pies"
],
"description": "John is a nice guy who likes pies.",
"married": false,
"name": "John",
"bookmarks": {
"HTTPie": "http://httpie.org",
}
}Please note that with this syntax the command gets unwieldy when sending complex data. In that case it's always better to use redirected input:
Submitting forms is very similar to sending JSON requests. Often the only difference is in adding the --form, -f option, which ensures that data fields are serialized as, and Content-Type is set to, application/x-www-form-urlencoded; charset=utf-8. It is possible to make form data the implicit content type instead of JSON via the config file.
POST /person/1 HTTP/1.1
Content-Type: application/x-www-form-urlencoded; charset=utf-8
name=John+SmithIf one or more file fields is present, the serialization and content type is multipart/form-data:
The request above is the same as if the following HTML form were submitted:
Note that @ is used to simulate a file upload form field, whereas =@ just embeds the file content as a regular text field value.
To set custom headers you can use the Header:Value notation:
GET / HTTP/1.1
Accept: */*
Accept-Encoding: gzip, deflate
Cookie: valued-visitor=yes;foo=bar
Host: example.org
Referer: http://httpie.org/
User-Agent: Bacon/1.0
X-Foo: BarThere are a couple of default headers that HTTPie sets:
GET / HTTP/1.1
Accept: */*
Accept-Encoding: gzip, deflate
User-Agent: HTTPie/<version>
Host: <taken-from-URL>Any of those—except for Host—can be overwritten and some of them unset.
To unset a previously specified header (such a one of the default headers), use Header::
To send a header with an empty value, use Header;:
The currently supported authentication schemes are Basic and Digest (see auth plugins for more). There are two flags that control authentication:
|
Pass a |
|
Specify the auth mechanism. Possible values are |
Authorization information from your ~/.netrc file is honored as well:
Additional authentication mechanism can be installed as plugins. They can be found on the Python Package Index. Here's a few picks:
- httpie-api-auth: ApiAuth
- httpie-aws-auth: AWS / Amazon S3
- httpie-edgegrid: EdgeGrid
- httpie-hmac-auth: HMAC
- httpie-jwt-auth: JWTAuth (JSON Web Tokens)
- httpie-negotiate: SPNEGO (GSS Negotiate)
- httpie-ntlm: NTLM (NT LAN Manager)
- httpie-oauth: OAuth
- requests-hawk: Hawk
By default, HTTP redirects are not followed and only the first response is shown:
To instruct HTTPie to follow the Location header of 30x responses and show the final response instead, use the --follow, -F option:
If you additionally wish to see the intermediary requests/responses, then use the --all option as well:
To change the default limit of maximum 30 redirects, use the --max-redirects=<limit> option:
You can specify proxies to be used through the --proxy argument for each protocol (which is included in the value in case of redirects across protocols):
With Basic authentication:
You can also configure proxies by environment variables HTTP_PROXY and HTTPS_PROXY, and the underlying Requests library will pick them up as well. If you want to disable proxies configured through the environment variables for certain hosts, you can specify them in NO_PROXY.
In your ~/.bash_profile:
Homebrew-installed HTTPie comes with SOCKS proxy support out of the box. To enable SOCKS proxy support for non-Homebrew installations, you'll need to install requests[socks] manually using pip:
Usage is the same as for other types of proxies:
To skip the host's SSL certificate verification, you can pass --verify=no (default is yes):
You can also use --verify=<CA_BUNDLE_PATH> to set a custom CA bundle path:
To use a client side certificate for the SSL communication, you can pass the path of the cert file with --cert:
If the private key is not contained in the cert file you may pass the path of the key file with --cert-key:
Use the --ssl=<PROTOCOL> to specify the desired protocol version to use. This will default to SSL v2.3 which will negotiate the highest protocol that both the server and your installation of OpenSSL support. The available protocols are ssl2.3, ssl3, tls1, tls1.1, tls1.2. (The actually available set of protocols may vary depending on your OpenSSL installation.)
If you use HTTPie with Python version lower than 2.7.9 (can be verified with http --debug) and need to talk to servers that use SNI (Server Name Indication) you need to install some additional dependencies:
You can use the following command to test SNI support:
By default, HTTPie only outputs the final response and the whole response message is printed (headers as well as the body). You can control what should be printed via several options:
--headers, -h |
Only the response headers are printed. |
--body, -b |
Only the response body is printed. |
|
Print the whole HTTP exchange (request and response). This option also enables |
--print, -p |
Selects parts of the HTTP exchange. |
--verbose can often be useful for debugging the request and generating documentation examples:
$ http --verbose PUT httpbin.org/put hello=world
PUT /put HTTP/1.1
Accept: application/json, */*
Accept-Encoding: gzip, deflate
Content-Type: application/json
Host: httpbin.org
User-Agent: HTTPie/0.2.7dev
{
"hello": "world"
}
HTTP/1.1 200 OK
Connection: keep-alive
Content-Length: 477
Content-Type: application/json
Date: Sun, 05 Aug 2012 00:25:23 GMT
Server: gunicorn/0.13.4
{
[…]
}All the other output options are under the hood just shortcuts for the more powerful --print, -p. It accepts a string of characters each of which represents a specific part of the HTTP exchange:
| Character | Stands for |
|---|---|
H |
request headers |
B |
request body |
h |
response headers |
b |
response body |
Print request and response headers:
To see all the HTTP communication, i.e. the final request/response as well as any possible intermediary requests/responses, use the --all option. The intermediary HTTP communication include followed redirects (with --follow), the first unauthorized request when HTTP digest authentication is used (--auth=digest), etc.
The intermediary requests/response are by default formatted according to --print, -p (and its shortcuts described above). If you'd like to change that, use the --history-print, -P option. It takes the same arguments as --print, -p but applies to the intermediary requests only.
As an optimization, the response body is downloaded from the server only if it's part of the output. This is similar to performing a HEAD request, except that it applies to any HTTP method you use.
Let's say that there is an API that returns the whole resource when it is updated, but you are only interested in the response headers to see the status code after an update:
Since we are only printing the HTTP headers here, the connection to the server is closed as soon as all the response headers have been received. Therefore, bandwidth and time isn't wasted downloading the body which you don't care about. The response headers are downloaded always, even if they are not part of the output
The universal method for passing request data is through redirected stdin (standard input)—piping. Such data is buffered and then with no further processing used as the request body. There are multiple useful ways to use piping:
Redirect from a file:
Or the output of another program:
You can use echo for simple data:
You can even pipe web services together using HTTPie:
You can use cat to enter multiline data on the terminal:
On OS X, you can send the contents of the clipboard with pbpaste:
Passing data through stdin cannot be combined with data fields specified on the command line:
To prevent HTTPie from reading stdin data you can use the --ignore-stdin option.
An alternative to redirected stdin is specifying a filename (as @/path/to/file) whose content is used as if it came from stdin.
It has the advantage that the Content-Type header is automatically set to the appropriate value based on the filename extension. For example, the following request sends the verbatim contents of that XML file with Content-Type: application/xml:
HTTPie does several things by default in order to make its terminal output easy to read.
Syntax highlighting is applied to HTTP headers and bodies (where it makes sense). You can choose your preferred color scheme via the --style option if you don't like the default one (see $ http --help for the possible values).
Also, the following formatting is applied:
- HTTP headers are sorted by name.
- JSON data is indented, sorted by keys, and unicode escapes are converted to the characters they represent.
One of these options can be used to control output processing:
|
Apply both colors and formatting. Default for terminal output. |
--pretty=colors |
Apply colors. |
--pretty=format |
Apply formatting. |
|
Disables output processing. Default for redirected output. |
Binary data is suppressed for terminal output, which makes it safe to perform requests to URLs that send back binary data. Binary data is suppressed also in redirected, but prettified output. The connection is closed as soon as we know that the response body is binary,
You will nearly instantly see something like this:
HTTP/1.1 200 OK
Accept-Ranges: bytes
Content-Encoding: gzip
Content-Type: video/quicktime
Transfer-Encoding: chunked
+-----------------------------------------+
| NOTE: binary data not shown in terminal |
+-----------------------------------------+HTTPie uses a different set of defaults for redirected output than for terminal output. The differences being:
- Formatting and colors aren't applied (unless
--prettyis specified). - Only the response body is printed (unless one of the output options is set).
- Also, binary data isn't suppressed.
The reason is to make piping HTTPie's output to another programs and downloading files work with no extra flags. Most of the time, only the raw response body is of an interest when the output is redirected.
Download a file:
Download an image of Octocat, resize it using ImageMagick, upload it elsewhere:
Force colorizing and formatting, and show both the request and the response in less pager:
The -R flag tells less to interpret color escape sequences included HTTPie`s output.
You can create a shortcut for invoking HTTPie with colorized and paged output by adding the following to your ~/.bash_profile:
HTTPie features a download mode in which it acts similarly to wget.
When enabled using the --download, -d flag, response headers are printed to the terminal (stderr), and a progress bar is shown while the response body is being saved to a file.
HTTP/1.1 200 OK
Content-Disposition: attachment; filename=httpie-master.tar.gz
Content-Length: 257336
Content-Type: application/x-gzip
Downloading 251.30 kB to "httpie-master.tar.gz"
Done. 251.30 kB in 2.73862s (91.76 kB/s)If not provided via --output, -o, the output filename will be determined from Content-Disposition (if available), or from the URL and Content-Type. If the guessed filename already exists, HTTPie adds a unique suffix to it.
You can also redirect the response body to another program while the response headers and progress are still shown in the terminal:
If --output, -o is specified, you can resume a partial download using the --continue, -c option. This only works with servers that support Range requests and 206 Partial Content responses. If the server doesn't support that, the whole file will simply be downloaded:
- The
--downloadoption only changes how the response body is treated. - You can still set custom headers, use sessions,
--verbose, -v, etc. --downloadalways implies--follow(redirects are followed).- HTTPie exits with status code
1(error) if the body hasn't been fully downloaded. Accept-Encodingcannot be set with--download.
Responses are downloaded and printed in chunks which allows for streaming and large file downloads without using too much memory. However, when colors and formatting is applied, the whole response is buffered and only then processed at once.
You can use the --stream, -S flag to make two things happen:
- The output is flushed in much smaller chunks without any buffering, which makes HTTPie behave kind of like
tail -ffor URLs. - Streaming becomes enabled even when the output is prettified: It will be applied to each line of the response and flushed immediately. This makes it possible to have a nice output for long-lived requests, such as one to the Twitter streaming API.
Prettified streamed response:
Streamed output by small chunks alá tail -f:
# Send each new tweet (JSON object) mentioning "Apple" to another
# server as soon as it arrives from the Twitter streaming API:
$ http --stream -f -a YOUR-TWITTER-NAME https://stream.twitter.com/1/statuses/filter.json track=Apple \
| while read tweet; do echo "$tweet" | http POST example.org/tweets ; doneBy default, every request HTTPie makes is completely independent of any previous ones to the same host.
However, HTTPie also supports persistent sessions via the --session=SESSION_NAME_OR_PATH option. In a session, custom headers—except for the ones starting with Content- or If-—, authorization, and cookies (manually specified or sent by the server) persist between requests to the same host.
All session data, including credentials, cookie data, and custom headers are stored in plain text. That means session files can also be created and edited manually in a text editor—they are regular JSON.
You can create one or more named session per host. For example, this is how you can create a new session named user1 for example.org:
From now on, you can refer to the session by its name. When you choose to use the session again, any the previously used authorization and HTTP headers will automatically be set:
To create or reuse a different session, simple specify a different name:
Named sessions' data is stored in JSON files in the directory ~/.httpie/sessions/<host>/<name>.json (%APPDATA%\httpie\sessions\<host>\<name>.json on Windows).
Instead of a name, you can also directly specify a path to a session file. This allows for sessions to be re-used across multiple hosts:
To use an existing session file without updating it from the request/response exchange once it is created, specify the session name via --session-read-only=SESSION_NAME_OR_PATH instead.
HTTPie uses a simple JSON config file.
The default location of the configuration file is ~/.httpie/config.json (or %APPDATA%\httpie\config.json on Windows). The config directory location can be changed by setting the HTTPIE_CONFIG_DIR environment variable. To view the exact location run http --debug.
The JSON file contains an object with the following keys:
An Array (by default empty) of default options that should be applied to every invocation of HTTPie.
For instance, you can use this option to change the default style and output options: "default_options": ["--style=fruity", "--body"] Another useful default option could be "--session=default" to make HTTPie always use sessions (one named default will automatically be used). Or you could change the implicit request content type from JSON to form by adding --form to the list.
HTTPie automatically stores some of its metadata here. Please do not change.
Default options from the config file, or specified any other way, can be unset for a particular invocation via --no-OPTION arguments passed on the command line (e.g., --no-style or --no-session).
When using HTTPie from shell scripts, it can be handy to set the --check-status flag. It instructs HTTPie to exit with an error if the HTTP status is one of 3xx, 4xx, or 5xx. The exit status will be 3 (unless --follow is set), 4, or 5, respectively.
#!/bin/bash
if http --check-status --ignore-stdin --timeout=2.5 HEAD example.org/health &> /dev/null; then
echo 'OK!'
else
case $? in
2) echo 'Request timed out!' ;;
3) echo 'Unexpected HTTP 3xx Redirection!' ;;
4) echo 'HTTP 4xx Client Error!' ;;
5) echo 'HTTP 5xx Server Error!' ;;
6) echo 'Exceeded --max-redirects=<n> redirects!' ;;
*) echo 'Other Error!' ;;
esac
fiThe default behaviour of automatically reading stdin is typically not desirable during non-interactive invocations. You most likely want use the --ignore-stdin option to disable it.
It is a common gotcha that without this option HTTPie seemingly hangs. What happens is that when HTTPie is invoked for example from a cron job, stdin is not connected to a terminal. Therefore, rules for redirected input apply, i.e., HTTPie starts to read it expecting that the request body will be passed through. And since there's no data nor EOF, it will be stuck. So unless you're piping some data to HTTPie, this flag should be used in scripts.
Also, it's might be good to override the default 30 second --timeout to something that suits you.
The syntax of the command arguments closely corresponds to the actual HTTP requests sent over the wire. It has the advantage that it's easy to remember and read. It is often possible to translate an HTTP request to an HTTPie argument list just by inlining the request elements. For example, compare this HTTP request:
POST /collection HTTP/1.1
X-API-Key: 123
User-Agent: Bacon/1.0
Content-Type: application/x-www-form-urlencoded
name=value&name2=value2with the HTTPie command that sends it:
Notice that both the order of elements and the syntax is very similar, and that only a small portion of the command is used to control HTTPie and doesn't directly correspond to any part of the request (here it's only -f asking HTTPie to send a form request).
The two modes, --pretty=all (default for terminal) and --pretty=none (default for redirected output), allow for both user-friendly interactive use and usage from scripts, where HTTPie serves as a generic HTTP client.
As HTTPie is still under heavy development, the existing command line syntax and some of the --OPTIONS may change slightly before HTTPie reaches its final version 1.0. All changes are recorded in the change log.
Please use the following support channels:
- GitHub issues for bug reports and feature requests.
- Our Gitter chat room to ask questions, discuss features, and for general discussion.
- StackOverflow to ask questions (please make sure to use the httpie tag).
- Tweet directly to @clihttp.
- You can also tweet directly to @jakubroztocil.
Under the hood, HTTPie uses these two amazing libraries:
HTTPie plays exceptionally well with the following tools:
- jq — CLI JSON processor that works great in conjunction with HTTPie
- http-prompt — interactive shell for HTTPie featuring autocomplete and command syntax highlighting
- httpcat — a lower-level sister utility of HTTPie for constructing raw HTTP requests on the command line.
- curl — a "Swiss knife" command line tool and an exceptional library for transferring data with URLs.
See CONTRIBUTING.rst.
See CHANGELOG.
BSD-3-Clause: LICENSE.
Jakub Roztocil (@jakubroztocil) created HTTPie and these fine people have contributed.