Area
Hardware-in-the-loop rigs and automated tests that run firmware on real boards, where an agent drives the test and reads the result.
The most active tools first.
25 d ago
espforge packages the ESP-IDF v5.5.4 framework and working method its author used to build an ESP32 weather display with Claude Code in 6 days. Its firmware components handle Wi-Fi setup and over-the-air updates with rollback. PC tools flash, probe, screenshot and test the board without a person. The README shows no recorded agent runs.
25 d ago
ae-hw-bridge connects AI agents over MCP to a target board through HW Puppet, MicroPython firmware for an ESP32-S3-DevKitC-1. Over one USB cable, HW Puppet drives the target's reset, boot and power lines and bridges its UART console. A mirrored terminal shows a person the same log the agent reads. Neither README shows a recorded run.
37 d ago
rttsh is a Windows command-line terminal over SEGGER J-Link RTT for checking firmware on the board. Firmware exposes a command console through RTT, and Lua scripts on the host send commands and assert on replies and timing. It shares the flashing probe, leaves the UART free and includes an 8-tool MCP server. The README's 23 example scripts target an STM32H743.
38 d ago
EmbeddedCI's BenchPod plugins connect Claude Code and Codex to a BenchPod, the company's hardware-in-the-loop tester wired to the board under test. One MCP server has tools to flash over SWD, power-cycle and capture the boot log, emulate I2C sensors, capture logic and analog signals and talk CAN. The vendor README lists the tools but shows no runs.
46 d ago
SAIHub firmware runs on an ESP32-C5 and serves MCP over Wi-Fi to coding agents such as Cursor and Claude Code. Agents can drive eight GPIO pins, generate PWM up to 50 kHz, capture edges, use a UART, switch 3.3 V and 5 V rails and run Lua scripts on the board. The README shows no photos or test results.
58 d ago
Agent Fingers runs on a stock Raspberry Pi Pico 2 as a signal instrument and CMSIS-DAP debug probe for AI agents. It offers eight digital channels, ADC capture, PWM, UART, I2C and SPI, plus SWD and JTAG for STM32 and ESP32 targets. Outputs need an expiring arm lease. The README shows no hardware runs.
Ranked by points weighed against age. New entries with early votes rise.
18 d ago
Andreas Spiess's 18-minute video shows Claude Code taking ESP32 firmware from a written specification to passing tests on a Raspberry Pi testbench. The agent writes and builds the firmware, flashes it to boards on the bench, runs tests over Wi-Fi, MQTT, BLE and RF, and fixes failures until the tests pass. The skills and testbench code are open source.
55 d ago
Tide & Wind Tracker shows tides and wind on an M5Stack PaperColor and tests its e-paper rendering on the ESP32-S3 itself. Test firmware renders fixtures sent over USB and returns the canvas bytes for comparison with golden images. Claude Code iterated against those tests while a person approved every change. The README has a photo of it running.
3 mo ago
On FastLED's unattended bench, an RP2040 stopped enumerating over USB after it was flashed with an I2S microphone sketch. Elevated device restarts and disable-enable cycles all failed. The issue says a port restart resets the bus but never cuts VBUS. The follow-up lists a switchable hub, an SWD probe and a watchdog that falls back to BOOTSEL.
4 mo ago
Martin Falch of CSS Electronics shows a Claude Code skill that decodes proprietary CAN signals into DBC files with a CANsub.2 interface. The skill matches bus bits against a reference, such as OBD2 data, dashboard video read by OCR or hand-logged gauge positions, and fits scale and offset. The 31-minute video runs it on car logs and a gauge bench.
4 mo ago
Zhihan Zhang and colleagues, mostly at the University of Washington, published Embedded Arena, a benchmark that scores agents' model and firmware edits on microcontrollers. Boards include a MAX78000, an ESP32-S3 and an STM32N6. The paper reports Claude Opus 4.7 and Gemini 3.1 Pro failing every deployment without hardware feedback; with it, the first success came within 3 iterations.
59 d ago
Sergio Marin built a flash lab in which an Orange Pi 3 LTS controls a Raspberry Pi Compute Module 5 for a coding agent. Through transistors it presses the power button and sets the boot mode, and it handles USB flashing and the serial console. He writes that clean flashes exposed a Wi-Fi bug a saved profile had hidden.
2 mo ago
Tulip Tree Technology has an AI coding agent write a hello-world for an STM32F030 and run it on Emerson, the company's own hardware emulator. The 8-minute video uses emctl to check that the firmware ran correctly, not just that it compiled. It also covers what Emerson models and what it leaves out. No physical board appears.
3 mo ago
UST announced an alliance with Anthropic that puts Claude into the engineering platforms UST runs for clients. In UST-iDEC, its hardware and silicon validation platform, Claude Code reads chip pinouts and schematics, then writes and runs regression test scripts that engineers used to write by hand. UST says the platform already cuts validation cycles by 50 to 70%.
3 mo ago
The FPGA Developer blog describes giving Claude Code control of programming, UART and power cycling on an AMD FPGA bench. In the post's example, the agent ports a reference design to a new dev board, then tests and fixes it there. The author says such ports take a few hours. The post also covers token use and hooks.
3 mo ago
HARDWARIO published a repository on 30 June that holds its TOWER IoT kit's protocol, STM32L083 firmware, CLI, flasher and hardware-in-the-loop bench. It began as a Claude Code control plane over submodules and is now a Rust monorepo with agent instructions. The README says bench runs flash a physical Core Module and measure power with a PPK2.
4 mo ago
Linutronix describes an off-the-shelf AI agent that tests embedded boards in its lab through a small MCP server. In the example session the agent reserves a board, flashes it, reads the console and power-cycles it after a bad flash. The post argues that the safety comes from the lab underneath. It shows one screenshot and no logs.
5 mo ago
Craig Schindler, CTO of Assemble Labs, describes how the company has AI agents write experiments for embedded devices and interpret sensor and power data. The 10-minute interview was recorded at NVIDIA GTC 2026. His main case study is a wearable health sensor, including how power regressions are found and fixed. The video shows no bench footage of the agents.
5 mo ago
rp2350-relay-6ch drives the six relays of a Waveshare RP2350-Relay-6CH board from Zephyr firmware, with a custom MCUmgr management group over USB CDC. Host tools include a Python RPC library, a CLI and a Linux daemon for test rigs. The author describes it as a first agentic coding project; the README has product photos only.
6 mo ago
An EEVblog forum poster says a hardware test on a self-built 68000 computer failed because GCC removed the write to address 0 as undefined behavior. Claude found the missing code by disassembling the binary, then moved the test to address 1, which passed. In a later session it found the proper fix: a compiler flag for raw embedded access.
14 mo ago
Embedder launched its firmware coding agent on Hacker News on 15 August 2025 with a demo adding a magnetometer to a Pebble 2 Duo. Founders Bob Wei and Ethan Gibbs said it reads uploaded datasheets and schematics, checks output over a serial console and steps through code with GDB. SEGGER announced J-Link and J-Trace probe access in September 2026.
15 mo ago
BootLoop, a Y Combinator Summer 2025 startup founded by Noah Pacik-Nelson and Chris Markus, is building an AI agent that writes and tests firmware. The company says the agent reads datasheets and schematics, then validates code on customer boards with oscilloscopes, logic analyzers and debug probes. In pilots, its own engineers run the agent and a person reviews every line.
