STM32Cube for Agentic AI Installer

The STM32Cube for Agentic AI installer is the command-line entry point for setting up and maintaining the STM32Cube Agentic Context toolkit on your machine and in your projects. Use cube agentic-ai for installer commands. Context bundle names follow the cube-agentic-ai-<connector> pattern.

This reference explains command syntax, defaults, deployment state, and project maintenance. For a first installation, use the installation guide in SEE ALSO. For context capabilities and typical development requests, use the STM32Cube for Agentic AI documentation in SEE ALSO.

The reference is published online and is also included in the installer bundle. Run cube agentic-ai get-user-manual --format html to open the bundled HTML manual, or use --format tldr for a concise command summary. The bundled manual describes the installed installer version. The online manual can describe a newer release.

It gives you a stable command set to:

  • Set up the toolkit in one command (setup): install on the machine, resolve what the contexts need, and deploy into a project.

  • Diagnose what the installed contexts need to work, and fix it (doctor).

  • Install STM32Cube Agentic Context bundles on your machine.

  • Deploy them into a project folder, wiring skills, agents, instructions, prompts, and MCP servers to your agentic AI system.

  • Lock the exact versions used in a project so the setup is reproducible.

  • Inspect what is installed and what is deployed at any time.

Command results use machine-readable JSON, so both human operators and agentic AI systems can drive the same workflow. Interactive commands also print prompts. Help and version requests print text.

Scope

The installer covers:

Area

What the installer does

Machine setup

Install context bundles via the cube bundle manager

Requirements

Check what contexts need (doctor), resolve dependencies, and report prerequisites and VS Code extension commands

Tool dependencies

When a context requires a Cube tool, the cube bundle manager installs it automatically — no separate action is required

Project deployment

Deploy STM32Cube Agentic Context files into your project folder

Target wiring

Place files in the right location for your IDE and agentic AI system, or use the portable .cube-ai/ layout

Version locking

Record exact context and tool dependency versions so projects are reproducible

Routing catalog

Generate a Markdown catalog from services declared by contexts deployed to the project

Contexts use consistent project-file ownership, so the installer can track the files it deploys for later updates and removal. When a context requires a Cube tool, the cube bundle manager installs that tool automatically.

The installer does not:

  • Install VS Code extensions without authorization: setup --install-vscode-extensions installs declared extensions, and doctor --fix --vscode-profile "<name>" installs missing declared extension dependencies only in that explicitly named, existing profile. Other extension checks report commands to run. The commands setup and doctor --fix also run a bundle’s declared Python-package installation. Neither installs Python itself.

  • Maintain a separate catalog of tools — tool lifecycle is the responsibility of the cube bundle manager via context dependencies.

  • Modify IDE settings beyond the documented project wiring and explicitly authorized extension installation.

Quick orientation

  • Concepts — concise conceptual model: context bundle vs tool bundle, automatic tool dependencies, versioning, scope boundaries, lock files, and routing catalog.

  • Prerequisites — what must be installed on your machine before running any command.

  • Commands reference — full syntax, options, and defaults for all commands.

  • Target selection — how --ide and --agentic-system interact, with the common deployment patterns.

  • Lock files — how the installer records deployed versions and service selections, and how to reproduce a project environment.

  • Project housekeeping — strategy guide for evolving deployments with add/remove/deploy flows.

  • Usage — annotated command sequences for the most common workflows.

  • Troubleshooting — error codes, their meaning, and the next action to take.

Routing behavior

Project commands use the usage-router mode by default. This mode generates the routing catalog from services declared by the contexts deployed to the project. It does not arbitrate overlapping context responsibilities. To opt in to overlap arbitration, pass --routing-mode usage-router-with-arbitration to a project deployment command.