Keysight Technologies is bringing agentic AI into radio-frequency engineering with new capabilities in its Advanced Design System (ADS) 2027 and RF Circuit Simulation Professional software. The technology allows engineers to direct AI agents through natural-language requests, while Model Context Protocol servers, reusable engineering workflows and Keysight simulation tools provide the context and validation needed to automate complex RF design tasks.
AI agents move into RF engineering
AI agents are increasingly being positioned as software workers capable of planning and executing multi-step tasks rather than simply generating text or code. But applying that model to engineering is more difficult.
In RF design, engineers work with schematics, layouts, simulation parameters and highly specialized workflows. Large language models can generate code and reason about technical documentation, but they do not inherently understand graphical engineering environments or the deterministic requirements of RF simulation.
Keysight Technologies is attempting to close that gap.
The company has introduced agentic AI capabilities into Advanced Design System (ADS) 2027 and RF Circuit Simulation Professional, allowing AI agents to interact with Keysight’s RF design and simulation software. Engineers can issue instructions in natural language, while the agent uses Model Context Protocol (MCP) servers to access documented tools and workflows within ADS.
The resulting design can then be evaluated through Keysight’s simulation environment.
That creates an important distinction from general-purpose AI coding assistants: the AI does not simply produce an answer and stop. It can execute engineering actions and use simulation to check whether the resulting design behaves as expected.
From AI-generated code to AI-executed engineering
The workflow is built around several components.
Keysight’s MCP servers provide agents with documented skills and tools for interacting with ADS. Engineers can also record workflows as macros, allowing established design practices to be captured and reused.
ADS can convert graphical designs into code that AI systems can work with, while Python script generation provides another mechanism for automating engineering operations.
An engineer could therefore describe a design task in natural language and have an agent perform a sequence of actions inside ADS rather than manually navigating every configuration and simulation step.
The approach addresses one of the biggest limitations of applying LLMs to specialized engineering software.
A language model may understand the concepts involved in RF design, but it cannot be trusted to independently infer every interaction with a complex EDA environment. By providing defined tools through MCP servers, the agent can execute specific operations using interfaces created and documented by the software provider.
That gives the agent a more deterministic execution layer.
Simulation becomes the AI’s engineering feedback loop
The second important component is validation.
After an agent completes a design task, Keysight simulation can evaluate the result. That creates a feedback loop in which AI-generated engineering work can be tested against technical requirements before engineers commit the design to physical hardware.
This is particularly important for RF systems because seemingly small changes can affect performance across frequency, power, impedance, noise and other characteristics.
Instead of asking an AI model whether a design is correct, engineers can use simulation to generate measurable evidence.
The distinction mirrors a broader movement in agentic AI: systems are becoming more useful when agents can interact with external tools that provide objective feedback.
In software engineering, that might mean compiling code or running automated tests. In RF engineering, the equivalent is simulation.
Keysight is effectively applying that principle to electronic design automation.
Agents can explore more design possibilities
One of the potential benefits is not simply faster execution, but a larger design search space.
RF engineers frequently spend time preparing simulations, configuring parameters, repeating tests and comparing alternatives. Automating those steps could allow an agent to explore more combinations within the same engineering cycle.
Keysight says the new capabilities can reduce repetitive setup and simulation work while allowing engineers to evaluate more scenarios.
The company does not claim that agents replace RF engineers. Instead, the workflow is designed to let engineers direct the AI while retaining simulation as a technical validation mechanism.
That distinction could become important as organizations adopt agentic engineering. The most practical systems may not be autonomous replacements for specialists, but supervised agents that automate portions of highly structured engineering workflows.
Engineering knowledge becomes reusable
Keysight is also using workflow capture as a mechanism for preserving engineering expertise.
Experienced engineers can record their methods as ADS macros. Those workflows can then be reused by colleagues and AI agents.
This creates a potential knowledge-management benefit beyond automation.
RF design organizations often depend on specialists who have accumulated years of experience around particular architectures, components and design practices. Capturing those processes in reusable workflows can make that expertise easier to distribute.
Keysight describes the longer-term opportunity as turning prior engineering work into organizational intelligence.
The concept is similar to the growing use of agent skills and tool libraries in enterprise AI: rather than asking an AI model to recreate a process from general knowledge each time, organizations provide reusable instructions and executable tools based on established practices.
MCP becomes an important bridge between AI and EDA
The use of Model Context Protocol (MCP) is also notable.
MCP is emerging as a standardized mechanism for connecting AI systems with external tools and data. In Keysight’s implementation, MCP servers provide AI agents with access to documented ADS capabilities.
That allows customers to use the LLMs and AI assistants they already work with instead of being tied to a single model.
It also reflects an increasingly important architectural pattern in enterprise AI. The model itself is becoming one component of a larger system consisting of context, tools, workflows, permissions and validation.
For specialized engineering applications, those surrounding components may be more important than the model’s ability to generate technically plausible text.
Agentic engineering reaches a more demanding test
Gartner expects more than 60% of organizations to deploy AI agents by 2028, according to the research cited by Keysight. The broader adoption trend is pushing vendors to demonstrate agentic AI in workflows where accuracy and repeatability matter.
RF engineering is one of those environments.
A chatbot can produce an imperfect answer without necessarily causing physical consequences. An incorrect RF design can result in failed simulations, engineering delays or expensive hardware revisions.
That makes deterministic tools and validation particularly important.
Keysight’s approach therefore represents a broader shift in AI engineering software: agents are being given controlled access to specialized tools and are being judged by the outputs those tools can validate, rather than by language generation alone.
Toward simulation-driven autonomous design
The new capabilities in ADS 2027 and RF Circuit Simulation Professional are available now, according to Keysight.
The longer-term opportunity extends beyond automating individual RF tasks. If agents can reuse engineering workflows, generate designs, run simulations and learn from validated outcomes, they could eventually participate in increasingly complex portions of the design cycle.
That does not eliminate the need for human engineering judgment. Instead, it changes where that judgment is applied.
Engineers could spend less time configuring repetitive simulations and more time defining objectives, constraints and trade-offs while agents explore possible solutions.
For RF and electronic design automation, that could be a more consequential use of AI than simply adding a chatbot to an engineering application.
Market Landscape
Agentic AI is moving from general productivity applications into specialized engineering environments, where the cost of incorrect outputs is considerably higher.
In EDA, AI is being applied to circuit design, verification, optimization, simulation and workflow automation. Keysight’s strategy is differentiated by combining LLMs + MCP tool access + reusable engineering workflows + simulation-based validation.
Competitors across the semiconductor and EDA ecosystem—including Cadence, Synopsys, Siemens and Ansys—are also developing AI-assisted engineering capabilities. The competitive advantage will increasingly depend on whether these systems can move beyond generating suggestions toward executing repeatable engineering tasks with measurable validation.
The broader trend is toward AI agents as engineering infrastructure rather than standalone assistants.
Top Insights
- Keysight is embedding AI agents directly into RF design workflows through ADS 2027 and RF Circuit Simulation Professional.
- MCP servers give agents structured access to documented engineering tools instead of relying solely on probabilistic LLM-generated instructions.
- Keysight simulation provides an objective validation layer for AI-generated RF designs before hardware implementation.
- Recorded engineering macros allow specialist workflows to become reusable assets for both human engineers and AI agents.
- Agentic RF design could increase the number of design scenarios engineers evaluate without requiring proportional growth in manual engineering effort.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
