diff --git a/Dockerfile b/Dockerfile
new file mode 100644
index 0000000..fb63bfd
--- /dev/null
+++ b/Dockerfile
@@ -0,0 +1,23 @@
+FROM python:3.6
+
+# prepare a build environment
+RUN apt-get update
+RUN apt-get install -y --no-install-recommends\
+ git bash gcc g++ gfortran libblas-dev liblapack-dev libpng-dev libffi-dev libfreetype6 libjpeg-dev libhdf5-dev
+
+WORKDIR /root
+# if we don't build numpy first scipy complains
+RUN pip install numpy
+RUN git clone https://github.com/openai/gym && cd gym && pip install .
+RUN git clone https://github.com/jvmancuso/ai-safety-gridworlds && cd ai-safety-gridworlds && pip install .
+RUN git clone https://github.com/david-lindner/safe-grid-gym && cd safe-grid-gym && pip install .
+RUN pip install tensorflow tensorboard torch
+
+COPY setup.py README.md run.sh main.py /root/
+COPY safe_grid_agents /root/safe_grid_agents
+RUN pip install .
+
+EXPOSE 6006
+
+ENTRYPOINT ["/bin/bash", "run.sh"]
+CMD ["-dc", "-E", "1000", "-V", "100", "-EE", "50", "corners",  "ppo-cnn", "-l", ".2", "-r", "32", "-e", "2", "-cc", ".2", "-eb", ".1"]
diff --git a/README.md b/README.md
index b3500aa..b23e73c 100644
--- a/README.md
+++ b/README.md
@@ -76,6 +76,18 @@ runs.
 Given a log directory `<logs>`, simply run `tensorboard --logdir <logs>`
 to visualize an agent's learning.
 
+## Docker
+
+Alternatively the agent can be ran in a docker container. To do this run
+
+``` {.bash}
+docker build . -t safe_grid_agents
+docker run -p 6006:6006 safe_grid_agents <args>
+```
+
+where `<args>` are the arguments you would normally pass to the main script.
+This does not use GPUs.
+
 # Development
 
 ## Code style
diff --git a/run.sh b/run.sh
new file mode 100644
index 0000000..487ed97
--- /dev/null
+++ b/run.sh
@@ -0,0 +1,4 @@
+BASE_DIRECTORY=`dirname "${BASH_SOURCE[0]}"`
+LOGDIR=$BASE_DIRECTORY/`hostname`-`date +%F_%H%M`
+python $BASE_DIRECTORY/main.py --log-dir $LOGDIR $@ &
+tensorboard --logdir $LOGDIR