node package manager
Share your code. npm Orgs help your team discover, share, and reuse code. Create a free org »



git push at this http router and it will host your branches on subdomains

build status

think bouncy + cicada



create an auth file and start the ploy server:

$ echo '{ "beep": "boop" }' > auth.json
$ sudo ploy server ./data -p 80 -a auth.json

then from a git repo with a server.js and/or a scripts.start in its package.json:

server.js should host its http server on process.env.PORT.

$ git remote add ploy http://beep:boop@localhost/_ploy/server.git
$ git push ploy master

Now your server.js will be running on http://localhost/. If you push again to master, in a few seconds the new master code will be running on http://localhost/.

To launch a staging instance on a subdomain, just push to a non-master branch:

$ git push ploy master:staging

Now go to http://staging.localhost/ to see your staging instance. (Edit /etc/hosts or set up dns wildcards with dnsmasq to test locally.)

Use ploy ls to list the running branches:

$ ploy ls


ploy does not detach your server processes. When the ploy server goes down, it takes the processes it started with it.

However, when the ploy server is started back up, it will attempt to restart all of the processes it was previously running.

When you git push code at a ploy server, your server will be started and any previous server running under the same branch name will be killed.


  ploy server DIRECTORY PORT
  ploy server { -d DIRECTORY | -p PORT | -a AUTHFILE | -f BOUNCER }
    Create a ploy http server, hosting repositories in DIRECTORY and listening
    on PORT for incoming connections.
    If AUTHFILE is given, it should be a json file that maps usernames to
    token strings to use for basic auth protection for ploy actions.
    Type `ploy help ssl` to show ssl options.
    You can specify a BOUNCER module with -f to use a custom http router.
  ploy ls { -r REMOTE | --verbose, -v | --format=FORMAT }
    List the running process branch names at REMOTE.
    Verbose formatting will use `branch,hash,repo,port`.
  ploy work { -r REMOTE | --format=FORMAT }
    List the working directories from checked out git commits.
    The default FORMAT is `repo,commit,time,branch`.
    Active branches are shown with a trailing *.
  ploy clean
    Remove unused working directories. These are the directories listed in
    `ploy work` without a *.
  ploy log NAME { -n ROWS | -f | -b BEGIN | -e END }
    Show ROWS of log output for the branch NAME like `tail`.
    Default -n value: screen height.
    Stream live updates when `-f` is set like `tail -f`.
    Slice log records for NAME directly with `-b` and `-e`.
  ploy log { -n ROWS | -f | -b BEGIN | -e END | --color=true }
    Show ROWS of log output for all branches.
    Lines will be prefaced with a colored branch name when stdout is a tty.
  ploy mv SRC DST { -r REMOTE }
    Move the branch name SRC to the DST branch name at REMOTE.
  ploy rm NAME { -r REMOTE }
    Remove the branch name at NAME, killing any running processes.
  ploy restart NAME { -r REMOTE }
    Restart the process at NAME.
  ploy stop NAME { -r REMOTE }
    Stop the process at NAME.
  ploy redeploy NAME { -r REMOTE }
    Redeploy the process at NAME, re-running install and test scripts.
  ploy help [TOPIC]
    Show this message or optionally a TOPIC.
    Topics: ssl
  For `ploy ls`, `ploy mv`, `ploy rm` commands that take a REMOTE parameter:
  REMOTE can be a git remote name or a remote URL to a ploy server. If there
  is exactly one ploy remote in set up as a git remote, it will be used by



ploy will look at your package.json's scripts.start field for how to start processes.

Before any services are started, npm install . will be run on the deployed repo. npm will handle the preinstall, install, and postinstall hooks. postinstall is a good place to put bundle steps like browserify.

If scripts.start is a string, ploy will set $PORT for a single process and host it accordinly.

If scripts.start is an object, the keys should map subdomains to commands to launch servers with. For instance:

  "scripts": {
    "start": {
      "beep": "node beep.js",
      "boop": "node boop.js",
      "index": "node server.js"

Will host beep.js at beep.domain, boop.js at boop.domain and server.js at just domain. When you push to non-master branches, domain will be prefaced accordingly to mount hosts at beep.staging.domain etc.

Each key can be a full url such as "" or just a subdomain.

Use the special key "index" to set a host to resolve for the root subdomain.

Each service start command will be immediately restarted when it crashes.


You can also have a scripts.stop that will have a $PID environment variable defined for the start script that was defined:

  "scripts": {
    "start": "sudo node server.js",
    "stop": "sudo kill $PID",

The scripts.stop keys work the same as the scripts.start keys:

  "scripts": {
    "start": {
      "x": "node x.js",
      "y": "node y.js"
    "stop": {
      "x": "node kill.js x",
      "y": "node kill.js y"


If you specify a scripts.test, it will be run.

If the test fails, code that you push will not be deployed and any old branches will continue to be run rather than be replaced.


You might want to manage non-http services with ploy. Just preface the services you don't want to show up in the http routing tables with an '_' like this:

  "scripts": {
    "start": {
      "web": "node web.js",
      "_notweb": "node notweb.js"

Underscored non-http services will not be given a $PORT to listen on so you'll need to figure out how you want multiple services running on the same box to work on your own. Consider registering your non-http services with seaport.


One handy trick you can do to make deploys much faster and more resilient against npm outages is to include a scripts.preinstall to copy over the node_modules directory from the previous build. This avoids a ton of extra network traffic and costly recompilation.

Here's an example scripts.preinstall:

  "dependencies": {
    "level": "^0.18.0"
  "scripts": {
    "start": "node server.js",
    "preinstall": "(cp -r $PREV/node_modules .; exit 0)"


Each script has these environment variables defined:

  • $COMMIT - git hash of the current deploy

  • $BRANCH - commit branch name, for example "master"

  • $REPO - repo name, for example "repo.git"

  • $PREV - the last previous working directory, if any

  • $PREV_COMMIT - git hash of the previous deploy, if any

  • $PREV_BRANCH - commit branch name of the previous deploy for the current $REPO, if any

Additionally, $PATH is prepended with node_modules/.bin so that bin scripts installed with npm will be available for install scripts.


var ploy = require('ploy')

var server = ploy(opts)

Create a new ploy instance, splitting opts between the underlying bouncy and cicada instances.

  • opts.repodir - directory to put git repo data
  • opts.workdir - directory to check out git repos into
  • opts.logdir - directory to store process stderr and stdout branch files
  • opts.auth - optional object mapping usernames to password token strings for basic auth
  • opts.bouncer - optional http router function. opts.bouncer(req, res, bounce) is called for each incoming request with a bounce() function from bouncy. To defer back to ploy, just call bounce() with no arguments.

If opts is a string, it will be used as the basedir for opts.repodir and opts.workdir.

The rest of the options will be looked over by bouncy to do things like set up an https server or whatever.

server.listen(opts, port, host...)

Call .listen() on the underlying http or https server, passing any opts object directly through to bouncy.

To host ploy over ssl, set the (opts.key,, and opts.cert), or set opts.pfx.

server.add(name, rec)

Add a a service under a branch name. rec should have:

  • rec.port - port where the http server lives
  • rec.hash - commit hash string
  • rec.process - process object to call .kill() on


Remove the process at the branch name, killing as necessary.


Restart the process at the branch name.

server.move(src, dst)

Move the process at branch name src to dst, killing the branch process at src if it was running.


Gather an array of the working directories in cb(err, working). Each w item in the working array has:

  • w.commit - the commit hash of the checked out working directory
  • w.time - the unix timestamp of when the commit was sent
  • w.branch - the branch name that was deployed
  • - the process id. undefined if inactive.
  • w.dir - the working directory
  • - whether the process is active


Remove inactive working directories.

server.redeploy(name, cb)

Re-run the deploy scripts for name at the current commit hash.


server.on('spawn', function (ps, info) {})

When a process is created from npm install or one of the package.json start scripts, this event fires with the info.commit and info.command executed under the info.env environment object.

server.on('output', function (name, stream) {})

When there is a new output stream for a branch, this event fires with the readable stream.

Output streams merge the output from all the processes used to start a branch.

running as non-root

On Linux you can easily setup port forwarding with iptables to avoid running ploy on port 80, which requires root privileges.

You'll need at least 2 new iptables rules to set it up. The first rule will redirect all incoming traffic from port 80 to port 8080:

iptables -A PREROUTING -t nat -p tcp --dport 80 -j REDIRECT --to-port 8080

The second rule will redirect all localhost traffic from port 80 to port 8080:

iptables -t nat -I OUTPUT -p tcp -d --dport 80 -j REDIRECT --to-ports 8080

You might also need a third rule if you'll push to ploy from localhost, but will use a hostname instead of localhost as ploy's remote url:

iptables -t nat -I OUTPUT -p tcp -d --dport 80 -j REDIRECT --to-ports 8080


With npm do:

npm install -g ploy

to get the ploy command or just

npm install ploy

to get the library.