What happened at MATLAB EXPO 2026 in Mumbai

The MATLAB EXPO 2026 conference, which MathWorks concluded on May 11, attracted over 1,300 participants. The main topics were agentic AI, digital engineering, and how these technologies are changing traditional development cycles — from the automotive industry and aerospace to chip manufacturing and medical devices.

At the conference, MathWorks demonstrated that agentic AI does not simply mean a chatbot inside MATLAB. It encompasses an entire ecosystem of tools that enable AI assistants to understand the context of engineering work, independently perform steps of the development cycle, and communicate with other tools across the team. Key innovations included Simulink Copilot, Polyspace Copilot, and expansions to tools such as the MATLAB Agentic Toolkit and MATLAB MCP Core Server, which MathWorks had already introduced as part of the April R2026a release.

"In engineering design and software verification, productivity gains cannot come at the expense of accuracy, traceability, or trust. MathWorks is committed to providing grounded AI tools for engineering that help teams accelerate their work while maintaining the discipline and confidence necessary for developing complex engineering systems," said Avinash Nehemiah, Head of Product Management at MathWorks.

Simulink Copilot: AI that understands your models

One of the most interesting innovations is Simulink Copilot — a generative AI assistant integrated directly into the Simulink environment used by engineers for Model-Based Design (design using block diagrams). This copilot is "grounded" in user models, defined processes, and MathWorks documentation — it does not hallucinate out of nowhere; its answers are based on the real context of the work the engineer is currently doing.

Simulink Copilot can generate model explanations, answer questions about their behavior, and help users find relevant blocks and subsystems. It can also isolate problems, propose solutions, and perform standardized tasks, speeding up design and increasing the consistency of development processes. For Czech engineering teams that use Simulink extensively — whether in automotive, automation, or energy — this means tangible time savings when debugging models and adhering to standards such as ISO 26262.

Polyspace Copilot: When AI checks code — even after AI

The second major innovation is Polyspace Copilot, which helps engineers interpret static code analysis results and more efficiently resolve identified issues. An interesting detail is that Polyspace as You Code — a new capability in R2026a — enables checking C and C++ code including code generated by AI tools. This means that if a developer uses ChatGPT or Claude to generate a piece of code, Polyspace can automatically check it for errors and security vulnerabilities.

For context: this capability is precisely what matters for industries where code must meet strict safety standards. In the automotive industry, this means ISO 26262; in aerospace, DO-178C; in medical devices, IEC 62304. MathWorks is sending the message: AI can write code, but it must also check it honestly.

Agentic AI workflows: MATLAB as part of a larger whole

What is perhaps most interesting, however, is how MathWorks is opening up MATLAB toward agentic workflows. Two key tools make this possible:

MATLAB Agentic Toolkit — an open-source toolkit (available on GitHub) enabling external AI agents to call MATLAB as a tool. In other words, an AI agent like Claude Code or ChatGPT can reach for MATLAB and run a simulation, data analysis, or optimization calculation, and use the result in subsequent steps.

MATLAB MCP Core Server — an implementation of the Model Context Protocol, a standard that allows AI models to communicate with external tools. Thanks to this server, MATLAB can function as a provider of computational and simulation services for any AI agent that supports MCP.

This means that MATLAB stops being an isolated tool and becomes part of a broader agentic ecosystem — much like Copilot in VS Code calling the terminal, but here an AI agent calls MATLAB. For Czech companies that use MATLAB (such as Škoda Auto, Honeywell, Thermo Fisher Scientific, or academic institutions), this opens the door to automating complex engineering tasks.

What else R2026a brought

In addition to copilots, R2026a also brings a host of other improvements. MATLAB Course Designer is a new product for educators that simplifies the creation of courses, exercises, and assessments using MATLAB and Simulink. Simulink FMU Builder enables the creation of standalone Functional Mockup Units — standardized models that can be integrated into various simulation environments.

The MATLAB application itself has also received enhancements, now allowing users to create and share interactive web pages with visualizations without requiring MATLAB to be installed. For developers working in Python, better Python environment management and data exchange between the two languages have been added. MATLAB Test can now generate starter tests and equivalence tests with the help of MATLAB Copilot.

In the area of signal processing, new Filter Designer and Filter Analyzer apps have been added, and a new Wireless Network Toolbox enables modeling and simulating wireless communication networks.

What this means for Czech engineers and companies

MATLAB is widely used in the Czech Republic at technical universities (CTU, BUT, UCT) and in industrial companies. The integration of agentic AI has several practical implications:

Shorter development cycles: Simulink Copilot can explain existing models to new team members in minutes instead of hours spent studying documentation. For rapid onboarding in the Czech automotive industry, this is a crucial advantage.

Fewer code errors: Polyspace Copilot combined with Polyspace as You Code checks even AI-generated code — meaning companies can experiment more safely with AI assistance when writing embedded code.

Greater automation: Thanks to the MATLAB Agentic Toolkit and MCP server, companies can build complex automation pipelines where an AI agent manages the entire process from design to verification.

The only downside for the Czech environment is that all these tools are available only in English — Czech is not yet supported. However, given that technical English is standard in engineering environments, this does not pose a significant barrier for the target audience.

MathWorks has not yet disclosed whether the pricing policy for these new AI tools will change. MATLAB Copilot is currently part of the standard MATLAB license; Simulink Copilot and Polyspace Copilot are part of R2026a. The price of a commercial MATLAB license starts at approximately $2,150 per year (about CZK 48,000), and Simulink is paid for separately — typically from $3,000 per year (about CZK 67,000) depending on the configuration.

Agentic AI: A trend transforming the entire industry

MathWorks is not alone in its push toward agentic AI. In recent months, we have seen a similar approach across the entire engineering software ecosystem — from Autodesk Fusion, where Anthropic integrated Claude for 3D modeling, to agentic AI in chip design, where tools from companies like Cadence and Synopsys are beginning to use AI agents for automating RTL design verification. According to EY analysis, agentic AI has the potential to boost global infrastructure productivity, and by 2050, an estimated $140 trillion in investments will be needed.

The next MATLAB EXPO — this time online — is scheduled for November 4–5, 2026. MathWorks is signaling that agentic AI and digital engineering are not just buzzwords for a single conference but a long-term strategy that will shape the future of engineering software.