Workbench for Zephyr (VS Code)
Ac6 Workbench for Zephyr is a VS Code extension that adds support of Zephyr development to Visual Studio Code, including host tools installation, SDK and toolchain management, west workspace management, project creation, build, flash and debug, Kconfig and devicetree editing, memory analysis, static code analysis and SBOM generation. The extension is available on the VS Code Marketplace and on Open VSX.
Features
Quick Start
Step-by-step guides are available for Windows, Linux and macOS. Supported toolchains
Zephyr SDKs installed globally with Supported debug runners
Debug sessions run on the built-in west debugserver backend or on Cortex-Debug. Runners and programming tools (including ModusToolbox Programming Tools) can be installed directly from the Install Runners view. Devicetree ManagerThe Devicetree Manager is a visual editor for the devicetree and the pin muxing of an application. It reads the resolved devicetree of the selected build configuration and generates the corresponding
AI agents (MCP)Workbench for Zephyr can be driven by an AI coding agent. The agent sees your applications and build configurations, runs builds in a VS Code terminal you can watch, and gets compiler errors back with the file and line, so it can fix them and build again. It can also read memory usage, Kconfig values and the resolved devicetree of a build, change a build configuration (board, overlays, conf files), change Kconfig options the way a build would accept them, search boards, shields, snippets and samples, and check what is missing from your environment. With a board connected, an agent can read what it prints on its serial console and send it a line, such as a Zephyr shell command. The capture shows live in a VS Code terminal you can stop, and a send asks you first. The capture keeps DTR and RTS low. Linux and macOS still raise both for a moment as a port opens, so a board whose reset is wired to them may reset once when a capture starts. An agent can also set things up from scratch: create an application from a sample or import one, change its toolchain, IntelliSense provider or Python environment, and run DT Doctor, the Kconfig hardening check, an SPDX SBOM or an ECLAIR analysis. With the full toolset it can also install and register toolchains (Zephyr SDK, Arm GNU, Rust, LLVM), create, update and edit west workspaces, and remove or delete what it no longer needs. When a choice is yours to make, such as an IAR licence token, it opens the right wizard or view for you instead. To connect an agent:
GitHub Copilot in VS Code 1.101 or later, and Cursor, need no setup at all. For any other MCP client, run Zephyr Workbench: Copy MCP Configuration for an Agent and paste the result into its configuration. One agent setup serves every project and every open VS Code window: each request goes to the window that has the application open. If an agent cannot reach Workbench for Zephyr, click Test connection in the AI Manager, or run Zephyr Workbench: Check AI Agent Connection (MCP). It runs the same command your agent runs, shows what to fix, and writes the full report to the Zephyr Workbench: MCP output. Everything stays on your machine: the server only accepts local connections, and each request needs a key that changes every time VS Code starts. Use the Permissions tab to choose what agents may do. With the Core preset, the default, agents build and inspect without asking, ask you before they create an application, change a west workspace, install a toolchain or send text to a board, and cannot delete anything. Full lets agents use every tool without asking, and Custom lets you choose Allow, Ask or Block for each tool. Only you can answer in VS Code. The AI Manager has two more pages. Zephyr Project MCP connects your agents, in the same way, to the Zephyr Project's own MCP server, run by Kapa.ai, which answers questions from Zephyr's documentation, source code and GitHub activity. It is a hosted service, so your questions leave your machine, and the first time an agent uses it you sign in once in your browser. A server you added to your Claude account as a connector shows as connected too: the page asks Claude Code which servers it reaches. The page also lists the sources it answers from and gives the configuration for any other MCP client. Third-party skills points to Zephyr agent skill collections written by others. CI and DockerA Docker image with VS Code and the Workbench for Zephyr extension pre-installed is available in the vscode-zephyr-workbench-docker repository. The image is based on Ubuntu 24.04 and already contains the Zephyr host tools. It can be used as a Dev Container from your local VS Code, as a standalone VS Code with GUI, or in terminal-only mode to run west builds in CI pipelines. DocumentationThe complete documentation is available on https://z-workbench.com/, in particular:
The latest features are listed in the CHANGELOG. Open SourceThe Workbench for Zephyr extension is a fully open-source project, built to provide an IDE for the Zephyr community and to introduce Zephyr to newcomers. The source code is available on GitHub, and we actively encourage developers to contribute, review and improve the project. The extension is licensed under the Apache-2.0 license. ContributeWe welcome contributions from developers of all skill levels. Whether you are fixing bugs, adding new features, or improving documentation, your efforts help make this extension better for everyone.
Useful links |
