Pow is a zero-configuration Rack server for Mac OS X. It makes
developing Rails and Rack applications as frictionless as
possible. You can install it in ten seconds and have your first app up
and running in under a minute. No mucking around with
compiling Apache modules, no editing configuration files or installing
preference panes. And running multiple apps with multiple versions of
Ruby is trivial.
How does it work? A few simple conventions eliminate the need for
tedious configuration. Pow runs as your user on an unprivileged port,
and includes both an HTTP and a DNS server. The installation process
sets up a firewall rule to forward incoming requests on port 80 to
Pow. It also sets up a system hook so that all DNS queries for a
special top-level domain (
.dev) resolve to your local machine.
To serve a Rack app, just symlink it into your
directory. Let's say you're working on an app that lives in
~/Projects/myapp. You'd like to access it at
http://myapp.dev/. Setting it up is as easy as:
$ cd ~/.pow $ ln -s ~/Projects/myapp
That's it! The name of the symlink (
myapp) determines the hostname
you use (
myapp.dev) to access the application it points to
Pow requires Mac OS X version 10.6 or newer. To install or upgrade Pow, just open a terminal and run this command:
$ curl get.pow.cx | sh
You can review the install script yourself before running it, if you'd like. Always a good idea.
The installer unpacks the latest Pow version into
~/Library/Application Support/Pow/Versions and points the
~/Library/Application Support/Pow/Current symlink there. It also
scripts for your user (the Pow server itself) and for the system (to
set up the
ipfw rule), if necessary. Then it boots the server.
Note: The firewall rule installed by Pow redirects all incoming traffic on port 80 to port 20559, where Pow runs. This means if you have another web server running on port 80, like the Apache that comes with Mac OS X, it will be inaccessible without either disabling the firewall rule or updating that server's configuration to listen on another port.
To install Pow from source, you'll need Node 0.6.0 or higher and npm:
$ git clone https://github.com/37signals/pow.git $ cd pow $ npm --global install $ npm --global run-script pow restart
For detailed instructions on installing Pow from source, including instructions on how to install Node and npm, see the Installation wiki page.
If you decide Pow's not for you, uninstallation is just as easy:
$ curl get.pow.cx/uninstall.sh | sh
Pow deals primarily with Rack applications. For the purposes of this
document, a Rack application is a directory with a
rackup file (and optionally a
public subdirectory containing static
assets). For more information on rackup files, see the Rack::Builder
Pow automatically spawns a worker process for an application the first time it's accessed, and will keep up to two workers running for each application. Workers are automatically terminated after 15 minutes of inactivity.
A virtual host specifies a mapping between a hostname and an
application. To install a virtual host, symlink a Rack application
~/.pow directory. The name of the symlink tells Pow which
hostname you want to use to access the application. For example, a
myapp will be accessible at
Note: The Pow installer creates
~/.pow as a convenient symlink
~/Library/Application Support/Pow/Hosts, the actual location
from which virtual host symlinks are read.
Once a virtual host is installed, it's also automatically accessible
from all subdomains of the named host. For example, the
virtual host described above could also be accessed at
http://assets.www.myapp.dev/. You can
override this behavior to, say, point
www.myapp.dev to a different
application — just create another virtual host symlink named
www.myapp for the application you want.
You might want to serve the same application from multiple hostnames. In Pow, an application may have more than one virtual host. Multiple symlinks that point to the same application will share the same worker processes.
If you attempt to access a domain that Pow doesn't understand, like
http://localhost/, you'll see a page letting you know that Pow is
installed and working correctly, with instructions on how to set up an
You can override this behavior to serve all requests for unhandled
domains with a particular Rack application. Create a symlink in
default and point it to the application of your
Pow's port proxying feature lets you route all web traffic on a
particular hostname to another port on your computer. To use it, just
create a file in
~/.pow (instead of a symlink) with the destination
port number as its contents.
For example, to forward all traffic for
$ echo 8080 > ~/.pow/proxiedapp
You can also use port proxying to access web apps written for other runtimes such as Python or Node.js. Remember that services behind the proxy won't automatically be started or stopped like Rack apps.
Sometimes you need to access your Pow virtual hosts from another computer on your local network — for example, when testing your application on a mobile device or from a Windows or Linux VM.
.dev domain will only work on your development
computer. However, you can use the special
domain to remotely access your Pow virtual hosts.
Construct your xip.io domain by appending your application's name to
your LAN IP address followed by
.xip.io. For example, if your
development computer's LAN IP address is
10.0.1.43, you can visit
myapp.dev from another computer on your local network using the URL
Pow lets you customize the environment in which worker processes
run. Before an application boots, Pow attempts to execute two scripts
.powenv — in the application's
root. Any environment variables exported from these scripts are passed
along to Rack.
For example, if you wanted to adjust the Ruby load path for a
particular application, you could modify
Pow supports two separate environment scripts with the intention that
one may be checked into your source control repository, leaving the
other free for any local overrides. If this sounds like something you
need, you'll want to keep
.powrc under version control, since it's
Pow invokes each application's Ruby processes in an isolated environment. This design makes it possible to use different Ruby runtimes on a per-application basis.
There are three ways to specify which version of Ruby to use for a particular application.
You can use rbenv to specify per-application Ruby versions for use with Pow.
rbenv local command lets you set a per-project Ruby version by
.rbenv-version file in the current directory. For example,
to configure a particular application to run with Ruby 1.9.3-p194,
change to the application's directory and run:
$ rbenv local 1.9.3-p194
Assuming you have set up rbenv in your login environment, there is no additional configuration necessary to use it with Pow.
For more information, see the rbenv documentation.
RVM is another option for specifying per-application Ruby versions for use with Pow.
You can create a project
file to specify an
application's Ruby version. For example, to configure your application
to run with Ruby 1.8.7, add the following to
.rvmrc in the
application's root directory:
Because RVM works by injecting itself into your shell, you must first
load it in each application's
.powenv file using the
if [ -f "$rvm_path/scripts/rvm" ] && [ -f ".rvmrc" ]; then source "$rvm_path/scripts/rvm" source ".rvmrc" fi
For more information, see the RVM web site.
You can adjust the
PATH environment variable in
.powenv to select Ruby versions on a per-application basis. For
example, if you have Ruby installed into
/opt/ruby-1.8.7, you can
add the following to
When Pow loads your application, it will find and use the first
binary in your
PATH — in this case
Pow automatically serves static files in the
public directory of
your application. It's possible to serve a completely static site
config.ru file as long as it has a
You can tell Pow to restart an application the next time it's
accessed. Simply save a file named
restart.txt in the
directory of your application (you'll need to create the directory
first if it doesn't exist). The easiest way to do this is with the
$ touch tmp/restart.txt
Restarting an application will also reload any environment scripts
.rvmrc) before booting the app, so don't
forget to touch
restart.txt if you make changes to these scripts.
It's also fine to kill worker processes manually — they'll
restart the next time you access the virtual host. A handy way to do
this is with OS X's Activity Monitor. Select "All Processes,
Hierarchically" from the dropdown at the top of the Activity Monitor
window. Then find the
pow process, expand the disclosure triangle,
find the Ruby worker process you want to kill, and choose "Quit
Process." (You can click "Inspect" on a worker process and choose
"Open Files and Ports" to determine which application the process is
It can be useful during development to reload your application with each request, and frameworks like Rails will handle such reloading for you. For pure Rack apps, or when using frameworks like Sinatra that don't manage code reloading, Pow can help.
tmp/always_restart.txt file is present in your application's
root, Pow will automatically reload the application before each request.
tmp/always_restart.txt will only reload the application,
not its environment scripts. To reload
.rvmrc, you must touch
Pow stores log files in the
~/Library/Logs/Pow directory so they can
be viewed easily with OS X's Console application. Each incoming
request URL is logged, along with its hostname and HTTP method, in the
access.log file. The stdout and stderr streams for each worker
process are captured and logged to the
apps directory in a file
matching the name of the application.
Note: Rails logger output does not appear in Pow's logs. You'll
tail -f log/development.log to see those entries.
Pow is designed so that most people will never need to configure it. Sometimes, though, you can't avoid having to adjust a setting or two.
When Pow boots, it executes the
.powconfig script in your
home directory if it's present. You can use this script to export
environment variables that will override Pow's default settings.
For example, this
~/.powconfig file tells Pow to kill idle
applications after 5 minutes (300 seconds) and spawn 3 workers per
export POW_TIMEOUT=300 export POW_WORKERS=3
See the Configuration class documentation for a full list of settings that you can change.
Note: After modifying a setting in
~/.powconfig, you'll need to
restart Pow for the change to take effect. See the Restarting Pow
POW_DOMAINS environment variable specifies a comma-separated
list of top-level domains for which Pow will serve DNS queries and
HTTP requests. The default value for this list is a single domain,
dev, meaning Pow will configure your system to resolve
127.0.0.1 and serve apps in
~/.pow under the
You can add additional domains to
If you want Pow to serve apps under additional top-level domains, but
not serve DNS queries for those domains, use the
variable. Entries in
POW_EXT_DOMAINS will not be configured with the
system resolver, so you must make sure they point to your computer by
Note: If you change the value of
POW_DOMAINS, you must reinstall
curl get.pow.cx | sh. This is because the relevant files in
/etc/resolver/ are created at install time.
WARNING: Adding top-level domains like ".com" to
be hazardous to your health! In the (likely) event that at some
point you lock yourself out of these domains, you will be unable to
reach important remote addresses like github.com (where you can find
the source code) and S3 (where Pow's installation and uninstallation
scripts are hosted). Do not panic! Delete the files Pow has created
/etc/resolver/ and DNS activity will return to normal. (You can
POW_EXT_DOMAINS for these domains instead.)
If you are writing software that interfaces with Pow, you can inspect
the running server's status and configuration via HTTP. To access this
information, open a connection to
localhost and issue a
request with the header
Host: pow. The available endponts are:
/status.json: A JSON-encoded object containing the current Pow server's process ID, version string, and the number of requests it's handled since it started.
/env.json: A JSON-encoded object representing the running server's environment, which is inherited by each application worker.
/config.json: A JSON-encoded object representing the running server's configuration.
Example of requesting an endpoint with
$ curl -H host:pow localhost/status.json
Alternatively, if you know the path to the Pow binary, you can
eval-safe version of the local configuration by invoking
Pow with the
--print-config option (useful for shell scripts):
$ eval $(~/Library/Application\ Support/Pow/Current/bin/pow \ --print-config) $ echo $POW_TIMEOUT 900
Pow runs as a Mac OS X Launch Agent. If the Pow server process terminates, the OS will restart it automatically.
You may need to manually restart Pow if you make configuration changes
~/.powconfig or modify your login environment. To tell Pow to
quit and restart, touch the global
$ touch ~/.pow/restart.txt
Alternatively, you can use the Activity Monitor application. Find the
pow process in the process listing, select it, and click the Quit
Powder is "syntactic sugar
for Pow." Install the gem (
gem install powder) and you'll get a
powder command-line utility that makes it easier to add
applications, tail your log files, and restart Pow. See the Powder
readme for more
examples of what it can do.
Powify is "an easy-to-use
wrapper for 37signals' Pow." Install the gem (
gem install powify)
to get a
powify command for installing, updating, and managing Pow
and your virtual hosts. See the Powify
readme for the full
list of commands.
Showoff is "the easiest way to share localhost
over the web." It creates a tunnel between your computer and the
Showoff server, then gives you a publicly accessible URL so others
can see the app or site you're working on. Showoff supports Pow's
virtual hosts — install the gem via
gem install showoff-io,
show dev. (with the trailing dot). Your Pow apps will be
If you're interested in contributing to Pow, first start by installing Pow from source.
Make your changes and use
cake to run the test suite:
$ cake test
Then submit a pull request on GitHub. Your patch is more likely to be merged if it's well-documented and well-tested. Read through the closed issues to get a feel for what's already been proposed and what a good patch looks like.
0.4.1 (May 16, 2013):
0.4.0 (June 7, 2012):
~/.powwith the virtual host as the filename and the port number as its contents, e.g.
echo 8080 > ~/.pow/myapp.
http://myapp.10.0.0.1.xip.io/(where 10.0.0.1 is your LAN IP address) instead of
.rvmrcfiles has been deprecated and will be removed in the next major release. See the "Specifying Ruby Versions with RVM" section of the manual for instructions on how to continue using RVM with Pow.
0.3.2 (August 13, 2011):
.rvmrcfile is present but rvm is not installed, Pow will fall back to whichever Ruby is in
$PATHrather than displaying an error message. This means you can migrate from rvm to rbenv without needing to remove
.rvmrcfrom your application.
0.3.1 (May 11, 2011):
POW_EXT_DOMAINSoption should actually work now. Apologies.
0.3.0 (May 10, 2011):
launchctl loadis invoked with the
tmp/always_restart.txtfile is present in an application, Pow will restart it before every request (useful with bare Rack apps or frameworks like Sinatra that don't manage code reloading).
POW_TIMEOUTis now specified in seconds instead of milliseconds, and a new
POW_EXT_DOMAINSoption defines a list of top-level domains that Pow will handle (but not serve DNS requests for).
Host:header no longer raise an exception, and URLs with
..are now properly passed through to the Rack application.
pow --print-config, and third-party apps can request information about the running Pow server via the
0.2.2 (April 7, 2011): Initial public release.
(The MIT License)
Copyright © 2012 Sam Stephenson, 37signals
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.