Agentic AI Automates Nanoscale Scientific Instruments with LLMs

Zahra Ayar, Marcos Penedo, Mahdi Mehdikhani, Nahid Hosseini, Prabhu Prasad Swain, Georg E. Fantner· August 28, 2026 View original

Key takeaways

  • Agentic AI can safely automate complex scientific instrument operation using LLMs.
  • The framework matches expert performance in image quality and tuning time for AFMs.
  • A guarded execution layer is crucial for preventing incorrect instrument commands.
  • LLM-based image assessment allows for adaptable strategies across diverse samples.

Who benefits

Scientific ResearchMaterials ScienceBiotechnologyPharmaceuticalsManufacturing

Summary

A new agentic AI framework uses a tool-augmented large language model to operate scientific instruments like atomic force microscopes. It automates command execution, image assessment, and post-processing, matching expert performance in live experiments.

This research introduces an agentic AI framework designed to automate the operation of complex scientific instruments, specifically demonstrated with an atomic force microscope (AFM). The system leverages a general-purpose, tool-augmented large language model (LLM) connected to instrument functions via a Model Context Protocol (MCP). It comprises three specialized agents: an AFM Messenger for command translation, an AFM Pilot for real-time image quality assessment and parameter adjustment, and an AFM Doctor for artifact diagnosis and post-processing. A key innovation is the LLM's ability to perform image assessment, allowing the same strategy to be applied across various sample types and imaging modes without specific retraining. The framework incorporates a crucial ambiguity check layer to ensure safe hardware operation, effectively eliminating wrong-command execution. Benchmarking against human experts showed the AFM Pilot matched their performance in image quality, iteration count, and tuning time, demonstrating a safe and effective path for AI-driven scientific instrument operation.

Why it matters

This development could significantly accelerate scientific discovery by automating tedious and expert-intensive instrument operations, freeing up researchers for higher-level analysis and experimental design.

How to implement this in your domain

  1. 1Evaluate current lab instrument workflows for repetitive or expert-dependent tasks suitable for AI automation.
  2. 2Investigate integrating LLM-based agent frameworks with existing instrument control software via APIs or custom protocols.
  3. 3Develop robust safety protocols and ambiguity checks to ensure secure and reliable autonomous operation.
  4. 4Pilot agentic AI on non-critical or simulated experiments to validate performance and refine parameters.
  5. 5Train and upskill lab personnel to oversee and collaborate with AI agents in advanced research settings.

Original post by Zahra Ayar, Marcos Penedo, Mahdi Mehdikhani, Nahid Hosseini, Prabhu Prasad Swain, Georg E. Fantner

"arXiv:2608.26198v1 Announce Type: new Abstract: Operating a scientific instrument such as an atomic force microscope (AFM) requires continuous expert decision-making. A trained user defines the experimental intent, translates it into instrument commands, assesses incoming data, a…"

View on X

Originally posted by Zahra Ayar, Marcos Penedo, Mahdi Mehdikhani, Nahid Hosseini, Prabhu Prasad Swain, Georg E. Fantner on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Engineering & DevTools