MC1391881 High
Microsoft Power Automate - Connect to local AI models from desktop flows
Summary AI-generated
Power Automate desktop flows can now connect to local AI models, enabling on-premises inference without sending data externally. Public preview starts June 30, 2026.
Written by Azure OpenAI (gpt-4.1) from the text of the post below. It can be incomplete or wrong; the original post is authoritative.
Similar posts
Search for more like this- MC1391903 Microsoft Power Automate - Run Python scripts using current Python versions in desktop flows
- MC1428573 Power Automate - Automate PowerPoint presentations in desktop flows
- MC1429626 Power Automate - Visualize desktop flows with flow chart view
- MC1411068 Power Automate – Build better forms with integrated Power Apps
- MC1411132 Power Automate – UI automation repair agent
- MC1383792 Microsoft Power Automate - Measure time and cost savings for desktop flows
Original post from Microsoft
We are announcing the ability to connect to local AI models from desktop flows in Microsoft Power Automate. This feature will reach public preview on June 30, 2026.
How does this affect me?
You can now add AI to your desktop flows without sending data outside your environment. By running models locally, you'll keep control of your data, meet privacy standards, and cut dependency on external services.
Users can connect to a locally hosted model by providing the endpoint URL, defining system and user prompts, and optionally include files as input. The action returns the model's response as an output variable, which can be used in subsequent steps of the desktop flow to drive decisions and enrich automation scenarios.
This capability supports models exposed through APIs compatible with the OpenAI standard, enabling integration with commonly used local model providers such as Ollama, vLLM, and LM Studio.
What action do I need to take?
This message is for awareness, and no action is required.
If you would like more information on this feature, please visit, please visit Connect to local AI models from desktop flows.
How does this affect me?
You can now add AI to your desktop flows without sending data outside your environment. By running models locally, you'll keep control of your data, meet privacy standards, and cut dependency on external services.
Users can connect to a locally hosted model by providing the endpoint URL, defining system and user prompts, and optionally include files as input. The action returns the model's response as an output variable, which can be used in subsequent steps of the desktop flow to drive decisions and enrich automation scenarios.
This capability supports models exposed through APIs compatible with the OpenAI standard, enabling integration with commonly used local model providers such as Ollama, vLLM, and LM Studio.
What action do I need to take?
This message is for awareness, and no action is required.
If you would like more information on this feature, please visit, please visit Connect to local AI models from desktop flows.