MBF Bioscience Is Adding Native Support for the Model Hardware Standard in ScanImage^® 

We’re adding native support for the Model Hardware Standard (MHS) in ScanImage as part of the MHS research preview. MHS is a new standard for AI agents to safely operate physical equipment in scientific research and advanced manufacturing. During the preview, we’re testing how MHS-compatible agents operate ScanImage within device-declared safety bounds.

 

Why we’re doing this

A two-photon experiment often involves several components: a microscope, behavioral hardware for stimulus and reward delivery, video systems, and other devices. Running a successful experiment requires careful coordination across all of them, from aligning imaging data with behavioral trials to synchronizing stimulus delivery and animal behavior.

 

ScanImage already helps solve much of this integration challenge on the acquisition side by controlling laser scanners, beam modulators, focusing devices, stages, and other microscope hardware from a wide range of vendors.

 

The remaining challenge is often connecting image acquisition with the rest of the experimental setup. Scientists frequently build these integrations themselves using custom scripts and workflows tailored to a particular experiment or lab.

 

MHS provides a standardized way to address that challenge. Each MHS-compatible device publishes its capabilities through a common interface, allowing AI agents and other software clients to interact with supported hardware in a consistent way.

 

Adding native support for MHS in ScanImage as part of the research preview is a step toward making complex experiments easier to coordinate and automate.

 

What it means in practice

Through ScanImage’s MHS interface, we’re testing how compatible AI agents and other software can interact with microscope acquisition capabilities controlled through vDAQ. 

 

• Set up and run an imaging session remotely. Configure scan geometry, timing, triggering, channels, and recording through the standard interface. 

• Read images as they are acquired. Access image data as it arrives, along with metadata identifying the region, channel, and acquisition timing. 

• Adjust a run in progress. Change a defined set of parameters mid-session without interrupting the scan. 

• Monitor the instrument. Track acquisition state, progress, and system health throughout the session, helping software distinguish a slow run from a failing one. 

 

Because MHS provides a standardized interface, different MHS clients can interact with the capabilities exposed by ScanImage through the same interface. 

 

Safety 

Devices declare bounds, interlocks, and emergency stops in the standard itself. AI agents inherit and operate within them by default. 

 

For ScanImage and vDAQ, that means: 

• A configuration is validated in full before an acquisition is armed. 

• The set of parameters that can change during a run is deliberately short. 

• An emergency stop reaches the hardware immediately, rather than waiting behind a queue of pending work. 

• The MHS interface will not operate unless ScanImage has been properly commissioned. 

 

We’re participating in the MHS research preview and exploring how AI agents can safely operate ScanImage. The ScanImage MHS driver is being tested as part of the research preview and is not yet a shipped feature or available to customers or partners. 

 

To learn more about MBF Bioscience’s work with MHS, contact support@mbfbioscience.com. 

 

For more information about the Model Hardware Standard and the research preview, read Anthropic’s Previewing the Model Hardware Standard announcement.