auto-embedded CLI
auto-embedded gives firmware projects five stages, a pin and DMA lock and a task log so the agent stops guessing.
- 262 ★ on the repository
- Last commit
Broad access
Why this level
- Flashing and debugging skills touch attached hardware and write to flash.
- Hooks alter the context of every request and run scripts in the project.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/auto-embeddedDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Flashing and debugging skills touch attached hardware and write to flash.
- Hooks alter the context of every request and run scripts in the project.
This is third-party code. Review the repository files before installing.
What it does
The `auto-embedded` framework installs into a firmware project with `aemb init` and connects rules to seven AI tools at once: Claude Code, Cursor, Codex, OpenCode, Copilot, Gemini CLI and Windsurf. Every agent reply names its stage: research, innovate, plan, execute, review. Code is written only after a plan a person has reviewed. Pins, DMA and interrupt priorities are first frozen in `hw-lock.yaml`, and a script catches conflicts by exit code. Progress and findings persist in `.auto-embedded/`, so a new session resumes the work. Inside are 24 skills for build, flashing, debugging, serial ports and buses, plus over 80 offline embedded articles.
Who it is for. Microcontroller developers who write STM32 and similar firmware with an AI agent and are tired of invented pins and registers.
Good fit when
- An SSD1306 driver is ported to STM32F103 after searching for an existing one.
- A USART1 interrupt does not fire and the cause is traced through registers and NVIC.
- A datasheet parameter is verified with a manual page reference.
- Several sessions work on one project and the context must not be lost.
Not a fit when
- The project has nothing to do with embedded systems.
- A commercial license is needed, since the source is under PolyForm Noncommercial.
- You are not ready to bind the agent to a five-stage process.
Example request
Port an SSD1306 driver to STM32F103: first look for an existing one in the knowledge base, freeze the pins in hw-lock.yaml and show the plan.Limitations
Requires Node.js 20 or newer and Python 3.9 or newer. The PolyForm Noncommercial license forbids commercial use, and the repository and README are mainly in Chinese, with an English version available. Flashing and debugging skills work with attached hardware, so review flash-write commands before they run. Hooks inject context into every request and subagent call, and those rules stay in the project after the agent is removed.
How to disable. Run `aemb uninstall` in the project, as described in the quick-start guide. Delete the `.auto-embedded/` folder and `hw-lock.yaml` by hand if unneeded.
FAQ
Which five stages are used?
Research, innovate, plan, execute and review. Every agent reply starts with a stage label, and code is written after the plan is approved.
What does hw-lock.yaml do?
It stores the frozen layout of pins, DMA and interrupts. On a conflict the script returns a non-zero exit code.
Which agents are supported?
Claude Code, Cursor, Codex, OpenCode, Copilot, Gemini CLI and Windsurf. Pick them with `--platforms` or install all through `--all`.
What persists between sessions?
Progress, changes and findings are written to `.auto-embedded/` and injected into context when a new session starts.
Is there a contest mode?
Yes. Contest mode runs six agent roles that first freeze pins and interfaces and then proceed in parallel.
Discussion
No questions or reviews yet.
Related
CLI tools24 skills for microcontroller firmware: build, flash, GDB debugging, UART, Modbus, CAN, logic analyzers and static analysis
A closed loop for STM32: the agent builds in Keil, flashes the board, reads serial output and finds the crashing line.
Claude Code skills and a Raspberry Pi testbench: the agent writes ESP32 firmware, flashes the board and tests it on hardware
Embedded Skills
embeddedskills
Skills for embedded development: AI drives compilers, debuggers (J-Link and others) and buses (CAN) from code to hardware checks. README in Chinese and English
ESP32 Workbench gives an agent 11 skills and 13 MCP tools for ESP32 firmware: pins, panics, build logs, partitions.
A terminal coding agent: plan mode before edits, settings import from Claude Code and Codex, skills, MCP, any model provider
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| auto-embedded (this card) | no | 262 | |
| Embedded Skills | no | 797 | embeddedskills collects skills for embedded work; auto-embedded adds a mandatory process, pin locking and task memory |
| Embedded AI Harness | no | 286 | embedded-ai-harness provides its own harness; auto-embedded wires rules into seven AI tools at once |
README in short
The five-stage rules and a resource table are installed into a firmware project by `aemb init`. Each platform's hooks inject the relevant context.