Software

SweepMe! brings AI agents to the measurement bench

Dresden-based SweepMe! GmbH has released version 1.6.1.9 of its measurement and automation software SweepMe!, which introduces the new Agentic Instrument Control tool. It lets AI agents operate lab instruments directly: users describe a measurement in plain language, and the agent runs the instruments, evaluates the data and delivers plots and key figures. Safety limits and control stay in SweepMe!, and with that, with the people at the bench.

Share this Post
With Agentic Instrument Control, SweepMe! serves as the link and the safety layer between the AI agent and the measurement bench. Graphic: SweepMe! GmbH

Contact info

SweepMe! GmbH

Bienertstraße 18
01187 Dresden

+49 351-41882423

https://sweep-me.net

contact@sweep-me.net

Contact person

Dr. Axel Fischer

Dr. Axel Fischer

An agent at the bench, not a script generator

Classic measurement automation works with fixed sequences that are programmed or configured in advance. Agentic Instrument Control takes a different approach. During a session, SweepMe! connects an AI agent to the instruments. The agent sends commands, reads measured values back in real time and decides on the next step based on the last result. This makes the tool particularly useful for bring-up and debugging, from ASICs to boards, wherever the next step depends on the last measurement.

An example from semiconductor characterization: the agent is asked to run a forward and reverse C-V measurement on a MOS capacitor at 100 kHz and to determine the oxide thickness, doping and flat-band voltage. It first performs an open and short correction on the LCR meter. It then contacts the sample on the prober, limits the bias voltage to ±3 V and runs the voltage ramp in both directions. Finally, it extracts the requested parameters, including the hysteresis.

Safety through fixed limits

Limits and permitted actions are stored in SweepMe!, not in the prompt. For each instrument, you can set a maximum voltage or a maximum current, for example. SweepMe! rejects values outside these limits. Users follow the agent’s work live, can stop it or take over at any time, and can switch between manual and AI control. Every command and every measured value is recorded. This allows sessions to be reviewed, handed over to colleagues and repeated.

From first prompt to repeatable test

Whatever works can be saved as a procedure. It measures, evaluates the result and reports whether the test has passed. Procedures can be re-run with a single click, with or without an agent, and combined into test flows that automate an entire test bench. On request, SweepMe! creates PDF reports with results and plots.

Open to any AI application

The agent connects through the SweepMe! MCP server (Model Context Protocol). Any AI application that supports MCP and tool calling can connect, including Claude, OpenAI Codex, Cursor, GitHub Copilot and Google Antigravity. For organizations with strict IT requirements, applications such as Msty Studio, AnythingLLM, LibreChat or LM Studio are a good fit, as they also work with locally hosted models. Because Agentic Instrument Control builds on the SweepMe! driver library, more than 300 instruments are supported. If a driver or a command is missing, the agent can write or adapt it. Every change can be accepted or reverted.

Availability

Agentic Instrument Control is part of SweepMe! Pro. When SweepMe! 1.6 is installed for the first time, SweepMe! Pro is available free of charge for one month. Trial licenses are also available on request. The SweepMe! AI Web Seminar on 26 November 2026 at 10:00 CET gives an insight into the AI features.

Further information: https://sweep-me.net/ai/agentic-instrument-control

Contact info

SweepMe! GmbH

Bienertstraße 18
01187 Dresden

+49 351-41882423

https://sweep-me.net

contact@sweep-me.net

Contact person

Dr. Axel Fischer

Dr. Axel Fischer