Hi  @<1544853721739956224:profile|QuizzicalFox36> , I think the correct method in that case is to run from docker. The agent handles the docker, the only thing is that you need to run the agent in docker mode using  --docker  . You can then then pass it in  docker_args  in  add_function_step  or even better you can add the entire script to the shell script that runs after docker init using  docker_bash_setup_script  also in  add_function_step
None
Thank you. I was considering this possibility but I hoped for another solution because this would require having multiple agents running on the same machine but in different docker containers for dfferent tasks. I couldn't find how to set up  clearml.conf  parameters  so that multiple agents could run  on same machine with different base images
 
				