Infineon adds hardware-aware AI to ModusToolbox

Infineon Technologies has introduced the ModusToolbox AI Assistant for Microsoft Visual Studio Code, supplying AI coding agents with information about the microcontroller, development board and software configuration used in an embedded project. The extension initially supports PSOC Edge and PSOC Control devices and is intended to reduce errors caused by generating code without sufficient knowledge of the target hardware.

Announced on 20 October, the extension analyses the project workspace locally and uses the resulting information to provide device-specific instructions to compatible AI assistants. The analysis covers the project structure, board support package, installed middleware, code examples and available development tools. It can also help establish the device, board and libraries required before code generation begins.

Embedded firmware depends on electrical and architectural details that cannot always be inferred from source code alone. Two microcontrollers using related processor cores may have different peripheral instances, interrupt arrangements, memory configurations and pin assignments. A programming interface available on one device may not exist on another, even when the application requirements appear similar.

General-purpose coding assistants can therefore generate plausible software using the wrong device family, an obsolete API or an unavailable hardware resource. Some faults become apparent when the compiler reports an undefined function, but others may survive compilation and emerge only after firmware is loaded onto the development board.

A peripheral initialisation sequence illustrates the problem. Configuring an SPI controller involves more than selecting a communications library. The firmware must use an available peripheral instance, configure the appropriate clock and pin functions, apply compatible timing parameters and coordinate transfers with the receiving device. An incorrect configuration can produce a successful build without establishing reliable communication.

The ModusToolbox extension supplies the AI agent with the target hardware context used to make those decisions. Its project analysis identifies relevant support packages and the software resources available in the workspace, allowing suggestions to be aligned more closely with the installed environment rather than with generic examples for the processor family.

Infineon identifies GPIO, I²C, SPI, UART and DMA among the supported development areas, alongside connectivity and real-time operating system functions. These interfaces have different electrical and timing requirements, so the value of project-specific assistance depends on whether the agent correctly uses the configuration information and applicable programming interfaces.

DMA is a demanding example because transfers must be coordinated with the source and destination, memory arrangement and triggering hardware. Firmware may need to initialise descriptors, manage buffer ownership and respond to completion events. Errors in these operations can cause data corruption or unexpected behaviour even when the main processor executes the surrounding application code correctly.

Real-time operating systems add scheduling considerations. Peripheral access may be shared between tasks, while interrupts can change the execution sequence at points that are difficult to reproduce during ordinary testing. A coding assistant aware of the installed RTOS and middleware can select more appropriate APIs, but it cannot independently prove that an application satisfies its worst-case timing or concurrency requirements.

Infineon also uses a planning step before code generation. The assistant creates a proposed project plan that defines the intended target device, board, libraries and project structure. Developers can review and amend that plan before proceeding, allowing unsuitable assumptions to be challenged before they are incorporated into the generated code.

The same workflow can support migration of existing projects between compatible Infineon devices. Such migration may involve changes to peripheral allocation, pin assignments and libraries, particularly where the destination MCU has a different collection of hardware resources. Reviewing those differences is more reliable than assuming that firmware can be transferred solely because both devices belong to the same semiconductor product range.

ModusToolbox tools including Device Configurator and Library Manager can be launched through the AI interaction. Device Configurator helps establish hardware resource settings, while Library Manager supports selection and management of project software dependencies. Connecting these tools to the development workflow allows configuration changes and code generation to use the same project context.

The extension is compatible with AI coding environments including GitHub Copilot, Cursor and Windsurf. Infineon states that its project-context analysis takes place entirely on the developer’s local machine, keeping the analysed workspace information within that system. This description concerns the extension’s local analysis; it should not be interpreted as a universal guarantee about the data-handling policies of every external AI service used alongside it.

Data confidentiality remains relevant when embedded software incorporates proprietary control algorithms, security functions or unreleased hardware details. Organisations must examine the configuration and policies of their chosen coding service, including whether prompts or generated responses are processed remotely. Local workspace analysis can reduce exposure within one part of the workflow without determining how every connected service handles information.

Generated firmware also remains subject to the limitations of the underlying hardware. The extension cannot create additional interrupt channels, memory capacity or analogue peripherals, nor can it establish that a proposed configuration satisfies the electrical requirements of external components. Correct code generation depends on the accuracy of the available project information and the assumptions supplied to the coding agent.

Infineon makes the assistant available as a Visual Studio Code extension pack. It requires editor version 1.85 and ModusToolbox 3.7 or later. PSOC Edge and PSOC Control are supported initially, while additions for PSOC 4, PSOC 6 and TRAVEO T2G are planned for subsequent releases. Those future device families should not be treated as supported by the initial version.

Normal firmware verification remains necessary after AI-assisted development. Compiler diagnostics and static analysis can identify some errors, while hardware testing establishes whether peripheral interactions and timing behave correctly on the actual board. Systems subject to security or functional safety requirements may also need traceable tests, design reviews and qualification evidence beyond the generated source code.

The ModusToolbox AI Assistant introduces a locally derived hardware context into the coding workflow, linking software suggestions with the actual devices and development resources selected for a project. Its effectiveness will be determined by the correctness of those suggestions and their behaviour under the same validation procedures applied to manually developed embedded software.