Keysight Advances its Design Software with Agentic AI

Keysight Technologies is opening a new route for artificial intelligence into radio frequency engineering. Announced on October 6, 2026, its latest software capabilities let engineers connect their own AI agents and large language models to Advanced Design System (ADS) 2027. Agents can help generate and optimize circuits, then use Keysight simulation to check their results. The capabilities are available in ADS 2027 and RF Circuit Simulation Professional.

The update combines Model Context Protocol (MCP) servers, workflow recording, and Python script generation. MCP servers give agents documented skills and tools built by Keysight that execute specific RF work consistently, reducing the variability of AI models generating answers statistically. Converting graphical designs into code gives agents a more accessible representation of engineering work. Engineers can issue natural language requests, with agents carrying out tasks through the software and running simulations to evaluate the output.

Why it matters begins with the difference between offering advice and performing work. A conversational assistant can describe a design procedure, but an agent connected to engineering software can execute steps and examine results. That creates the possibility of a shorter loop between an engineer’s intention, a proposed circuit, and evidence about its performance.

Consider a hypothetical amplifier project. An engineer might ask an agent to explore component variations against specified performance targets. If the workflow supports those operations, the agent could organize repeated simulations and return candidates for review. The engineer would still decide whether the targets, assumptions, and resulting tradeoffs make sense. The potential gain is more time for judgment and less time spent operating menus or assembling repetitive scripts.

That distinction matters because faster execution can expand the search for better designs. Teams working against a deadline often have to choose how many alternatives they can investigate. If automation lowers the effort required for each experiment, engineers could evaluate additional configurations before committing to hardware. The benefit would depend on whether those experiments answer useful questions, rather than simply producing more results.

Simulation provides an important checkpoint, but its role needs careful interpretation. An AI system’s plausible explanation does not establish that a circuit meets its requirements. Evaluating the proposed design through engineering tools introduces evidence beyond the model’s language output. However, a successful simulation remains conditional on the models, constraints, and test conditions supplied. Engineers still need to examine whether the evaluation represents the intended application.

Another implication concerns institutional knowledge. Experienced designers accumulate practical methods that may be difficult to transfer through documentation alone. Recording a workflow could preserve some of that procedural knowledge in a form colleagues can execute and adapt. Over time, reusable procedures could help teams maintain continuity when specialists change roles or support several projects.

Keysight’s approach also builds on an existing automation foundation. Its earlier Python automation materials describe using APIs for enterprise workflows, external optimization, and AI applications. In December 2025, the company announced ADS Chat and Copilot assistants for guidance and command execution. Viewed against that background, the latest announcement suggests a progression toward agents coordinating broader engineering tasks.

The open connection to customers’ chosen AI systems could also influence adoption. The MCP servers work with the AI assistants and LLMs organizations already use, and with tools from multiple vendors in the same workflow. Organizations may want engineering tools to fit their existing AI arrangements and workflows. An interoperable interface could make experimentation easier, although teams would still need to establish appropriate access, protect proprietary designs, and track automated changes.

Bottom line: For engineering leaders, the practical opportunity is to test bounded workflows with clear success criteria: time saved, useful alternatives explored, reproducibility, and review effort. If those trials succeed, agentic AI could become a valuable extension of the design team, helping engineers turn more ideas into evaluated options while retaining responsibility for the decisions that ultimately shape hardware. That is where its lasting engineering value will emerge: in better decisions supported by repeatable, inspectable work.

Resources

• Web Page: Advanced Design System (ADS)
• Web Page: RF Circuit Simulation Professional (Nexus)
• White Paper: Planning Your AI Design Journey in RF EDA
• Case Study: Custom End-to-End AI System Multiplies Productivity and Boosts RFIC Performance
  
• Webinar: How to Capture and Orchestrate a Power Amplifier Design Workflow

Also Read:

Podcast EP367: How Keysight is Accelerating RF and Microwave Design with Matthew Ozalas

Keysight EDA update at DAC 2026

DAC 2026: A Discussion of Who Owns the Intelligence Behind Tomorrow’s Chips and How to Make it Better