DEV Community

Emily Cain
Emily Cain

Posted on

How to Set up a Twitter Bot with Python and Heroku

I just set up a Twitter bot to generate pharmaceutical ads and tweet them out a few times a day. (Source code here). I ended up piecing together concepts from a couple different tutorials in order to do this, and no single tutorial laid out the whole process, so I'm writing it up here.

image description: xkcd comic 1988. Transcript can be found at

The steps can be summarized as:

  1. write a function to generate text strings for your bot to tweet
  2. create a Twitter account, and set up an application under it
    • set up a phone number with Google Voice
    • assign your account that phone number, allowing you to create applications
  3. write a script that connects to your Twitter application, and sends it an API call to tweet the output of your function from step 1
  4. set up a Heroku app that runs the function from step 3 as a worker
  5. set up a basic server in your Heroku app so that Heroku won't freak out and throw an error
  6. set environment variables in Heroku for your Twitter app's credentials.

Steps 4 through 6 could also be used to set up any Python script to run on Heroku.

I'm using some Google features to generate a unique email and phone number for the Twitter account, but you can handle creating those with whatever service you like.

This article assumes some basic background knowledge on the following concepts: Python, REST APIs, Git, environment variables.

Because it uses Heroku, the app will have to be saved in a Git repository.

Also, I used a virtual environment and saved the requirements in a file by running

$ pip freeze > requirements.txt
Enter fullscreen mode Exit fullscreen mode

This becomes relevant when you set up Heroku, which uses requirements.txt to install the Python packages you need.

1. Generate strings

I'd made this functionality a while back. For the pharma bot, I use Markov chains to generate drug names, and then insert them in a randomly generated template, to get advertisements like this:

Worried you might have twitchy ossicle? Sudate® can help!

Are you plagued by restless elbow? Ask your doctor about Estron® today!

You don't have to deal with pompous spleen alone. Ask your doctor about Ryzema Prenztriumetine® today!

The function that creates these is called get_ad.

In your own code, replace get_ad with a function that generates whatever kind of text you like.

2. Create a Twitter app

You'll probably want to give the bot its own account rather than Tweeting from an existing account, so set one up.

The Twitter account will need a unique email. If you want to use an email address that already has its own Twitter account, you can take advantage of Gmail's functionality to sign up with an "alternate" email that goes to the same inbox.

You will also need to set a phone number in your app; again, you can use Google to create an alternate contact that goes to an existing number, this time using Google Voice.
Next, go to your Twitter Apps dashboard and create a new app (while logged in as the bot's account).

a screenshot of the Twitter application creation page

For the app's website, I put the GitHub URL where the repo for this project lives.

Once you create the app, make sure that it allows read and write access.

3. Connect to the Twitter App and send tweets

I used this article, How To Write a Twitter Bot with Python and tweepy, as a rough guide for this part. Check it out for a guide on setting up your Twitter app and connecting to it via API. Note that you will need to add an "access key" to your Twitter application.

They have their bot reading lines from a text file; instead, I had mine call get_ad. The script that does this is called Here's the part that generates and sends the Tweet:

from generate_advertisement import get_ad

INTERVAL = 60 * 60 * 6  # tweet every 6 hours
# INTERVAL = 15  # every 15 seconds, for testing

auth = tweepy.OAuthHandler(CONSUMER_KEY, CONSUMER_SECRET)
auth.set_access_token(ACCESS_KEY, ACCESS_SECRET)
api = tweepy.API(auth)

while True:
    print("about to get ad...")
    ad = get_ad()
Enter fullscreen mode Exit fullscreen mode

I then ran in my local terminal.


screenshot of initial tweets sent by bot

It was able to send tweets just fine, but I don't want to leave my laptop on and my terminal open forever, so I needed to get it running on its own server somewhere. My next task was to put the bot on Heroku.


Connecting to Twitter via the API (including via a wrapper library like tweepy) requires you to use keys that identify your app and authenticate you as a user. Make sure you don't commit these keys in a Git repository--that's like publishing your password. Even if you later remove the keys from your code, they'll be in your commit history.

Initially, I stored the keys in a file called, which I added to my .gitignore. It looked like this:

CONSUMER_KEY = 'consumer key'
CONSUMER_SECRET = 'consumer secret'
ACCESS_KEY = 'access key'
ACCESS_SECRET = 'access secret'
Enter fullscreen mode Exit fullscreen mode

I then imported the keys into This didn't end up working out once I put the app on Heroku, but it was fine for testing it on my local machine.

4. Set up Heroku

Sign up for a Heroku account if you don't already have one. Heroku allows you to push and deploy code with Git, and you can create up to 5 apps for free.

Once you have the Heroku command line interface installed, initialize your repository with Heroku. It's ok if you let it generate a random URL for your app, since nobody will be looking at it directly.

$ heroku create
Enter fullscreen mode Exit fullscreen mode

This will set up a Heroku app and also give your Git repository a remote called heroku. The code on that remote's master branch will be run by Heroku.

You'll need to set up a Procfile to tell Heroku what to do once it gets your code. Put this in your Procfile:

worker: python
Enter fullscreen mode Exit fullscreen mode

This means Heroku will do the equivalent of when you ran python on your local terminal.

Then commit your code and push it to Heroku.

To deploy from your local master branch:

$ git push heroku master
Enter fullscreen mode Exit fullscreen mode

To deploy from a different branch:

$ git push heroku my_branch_name:master
Enter fullscreen mode Exit fullscreen mode

Then you can log into your Heroku dashboard and check its activity and logs to see what happens.

5. Basic Web Server

Turns out Heroku doesn't like to just run scripts with workers, because it expects to see a web server.

You can fix this by creating a simple web server. I used this StackOverflow answer as a guide.

Code (

from os import environ
from flask import Flask

app = Flask(__name__)'PORT'))
Enter fullscreen mode Exit fullscreen mode

Then, in your Procfile, just say: web: python

I tried this (and committed and pushed to Heroku again). I then saw the following error:

Error R10 (Boot timeout) -> Web process failed to bind to $PORT within 60 seconds of launch python flask
Enter fullscreen mode Exit fullscreen mode

I fixed this by setting the host when I ran the app: '', port=environ.get('PORT'))
Enter fullscreen mode Exit fullscreen mode

6. Set environment variables

Remember how I said not to commit your Twitter credentials to your Git repository? This becomes an issue when you deploy to Heroku, because it uses your Git repo to deploy, and so importing constants from an ignored file won't work.

You can set environment variables in your Heroku dashboard, then use os.environ to get them in Python. Add this to before making the API call:

from os import environ
Enter fullscreen mode Exit fullscreen mode

Then your app will be able to connect to the Twitter API without ever committing your secrets.

What's Next?

Now you know how to create a Twitter bot and run it on Heroku!

You can adapt this project to tweet out anything you like. This article explains how to use Python to generate text with Markov chains. You can create a text file with each line containing a sentence you want to tweet using the method outlined in How To Write a Twitter Bot with Python and tweepy. You can tweet lines from public domain works at Project Gutenberg. Or do anything you want!

note: an earlier version of this article was published at

Top comments (15)

Sloan, the sloth mascot
Comment deleted
emcain profile image
Emily Cain

Can you elaborate? What step are you on? What have you done so far?

Sloan, the sloth mascot
Comment deleted
emcain profile image
Emily Cain

The code on my GitHub is the most recent version.

What happens when you try to run it in the command line (commenting out the env variable loading and uncommenting the import from constants)?

wolverine2k profile image
Naresh Mehta

Heroku plus uptime or cronjob can help you automate the bot. My bot is up and running @ My blog post below has more details.

narayanbhat166 profile image

Hey my python worker just crashes everytime i try to run it
i get "state changed from starting to crashed"
and returns status code 0
but in my program i send a telegram message, i dont receive any message

zogzmiller profile image

I recently tried following this article to set up a twitter bot that replies with sentiment analysis of whatever user the person requests. Currently the bot works and runs through terminal but is not doing anything once deployed in heroku. If someone is willing to take a look at what I have so far and give feedback would be greatly appreciated. The github link below has my heroku bot folder

jordytrevor profile image
Trevor Jordy

Hi, I just followed your guide and pushed everything successfully, but nothing is happening. How can I tell if my bot is running properly? I have print() statements that I should be seeing in the logs. The bot builds correctly its just nothing happens. Does it run automatically, or do I have to do something to run it?

emcain profile image
Emily Cain

I just saw you have a Twitter account hooked up to it? Is it publishing tweets?

cokastefan profile image
Stefan Petrushevski

You can also use heroku scheduler without setting up a web app. Just set it up to run python at your desired interval.

alialhajji profile image
Ali Alhajji

Good job. I've been looking for such a solution to run my code other than my local machine. Thnx Emily!

willemm6 profile image

Nicely done!

meronzonzo profile image

Good post! I did the same for a bot I just rewrote (@linuxsecbot). I was about to setup flask when I saw that Heroku supports scheduled tasks. With python is done with APScheduler. Thanks for posting!

sheenbanach profile image

wow this is an awesome post thanks

aakash profile image
Aakash Deep

great post! thanks.
I've a question though, how do I see the logs from the python script? executing 'heroku logs' gives me the logs related to the web server.