Hi everyone,
I’m trying to add an OkPy script to an Orkestra Batch Preset, but the script doesn’t appear in the list of available actions.
The script is published, visible in Orkestra, and runs manually without any issues. However, it is not available when configuring a Batch Preset.
Are OkPy scripts supported in Batch Presets, or are there any additional requirements/settings needed for them to appear?
Thanks in advance for any insights.
Hey there @AdrianGrosu !
You are correct that batch runnin OkPy tools is currently not available yet, because their (often) rich UIs induce the need for a preset system so they can run unattended. This is on the roadmap and planned for a release before end of 2026 (i’ll update this thread 
) .
In the meantime! If what you’re after is being able to run this tool on multiple models , OkPy Agent can make it possible ! All you’d have to do is open that tool and ask the agent : “This tool works well, but I need to be able to run it on multiple models , including ACC cloudshared models” .
The agent will adress all the batching capatibilites! The only thing that would be missing compared to Orkestra Batch Presets is ability to schedule .
Let me know how it goes!!
Hi @m.elayoubi
Thanks for the prompt clarification! I can’t wait to test this feature once it becomes available. It would definitely make workflows like this much easier to scale across multiple projects.
In the meantime, I’ll give the suggested OkPy Agent approach a try and see how it performs on a few test models. It sounds like a promising alternative while we wait for batch support.
One question I do have is around data privacy and security. I’m generally cautious about connecting AI agents to live project models, especially on client projects. Could you share a bit more about what model data the agent can access, what information is sent externally (if any), and how confidentiality is handled when working with live Revit or ACC-hosted models? if this was already explained anywhere please share.
Thank you! I’m looking forward to testing the new functionality when it becomes available
of course!
The agent works inside your own Revit session with your existing permissions, and the tools are deterministic code. So once you’ve tested a tool and trust it, running it on multiple files is very reliable. The Agent doesn’t “look” at your ACC models, it just has a way to list them for your to pick and run your automations on. It uses your logged in revit session to list files the same way revit does.
What is sent is only the conversation (your prompts, the generated code..). It goes encrypted through Orkestra’s server , and currently to Anthropic’s commercial API (no data retention and no training on data). Workstations never call an AI provider directly.
The agent writes code so nothing changes your model unless you run it. If you are on super sensitive projects and don’t want any parameter value or family names etc. (the type of info that can end up in context), one safe option is to develop and test your tools on sandbox projects, then use them on the real project. Using the tools calls no AI whatsoever.
Another option we do offer, but is for very large enterprise deal and not availalble on regular subscription, is to host our finetuned model on your own company hardware (or rented instances) and we route OkPy Agent there, makinig100% sure your data never leaves your network and never reaches any AI providor.
Thank you @m.elayoubi for the detailed explanation and for taking the time to clarify how everything works.
This definitely addresses my main concerns around data privacy and model access. I think I’ll follow your recommendation and develop/test any new tools on sandbox projects first, then deploy them on live projects once they’re proven and trusted.
I’ll also keep this in mind if my current company decides to scale the use of AI-driven automation in the future. It’s good to know there are enterprise options available for organizations with stricter security requirements.
In any case, I have to say what you’ve built here is genuinely groundbreaking and quite disruptive to the way we traditionally approach automation.
Really impressive work by the whole team. I’m looking forward to experimenting with it further and seeing these capabilities continue to evolve. I’ll definitely give the Agent approach a try and will keep in touch if we decide to scale this more broadly within the company. Thanks again! KR Adrian
Thanks so much @AdrianGrosu !! 
means to world to us!!
It only gets better from here too!
