v0.4.2 · local-first · macOS · Linux · Windows

AI coding tools forgot
about hardware.
We didn't.

InteJer is the desktop AI agent built for embedded, IoT, and connected-device engineers. It reads datasheets. It drives debuggers. And it builds the entire IoT stack with you — from the hardware layer up to the frontend your customers actually see.

Download for macOSWatch the 90-second demoalso: .exe · .deb · .AppImage
Built for these workflows
ESP32
ESP32
STM32
STM32
RP2040
RP2040
nRF
nRF
ARDUINO
ARDUINO
ZEPHYR
ZEPHYR
FREERTOS
FREERTOS
ROS
ROS
…and any board with a PlatformIO target.
What InteJer can do

One agent. Your whole hardware stack.

Five capabilities you won't find anywhere else. Each one shipping today.

DOMAIN-NATIVE

Knows ESP32, STM32, Zephyr out of the box.

Skill packs for arduino-development, esp32-development, stm32-development, zephyr-rtos-development — auto-loaded by the agent when the conversation matches the domain.

DATASHEET-AWARE

From part number to PDF in one chat.

partfetch resolves LM358 → datasheet, parsed pinout tables, embedded images, and freely-linked CAD (STEP / IGES / STL).

DEBUG-PROBE READY

Your AI can set a breakpoint on your STM32.

22 tools over probe-rs. Flash, halt, step, read registers and memory. J-Link, ST-Link, DAPLink, CMSIS-DAP. ARM Cortex-M and RISC-V.

IOT-CLOUD CONNECTED

First-class MCPs for the IoT stack.

ThingsBoard device management, MongoDB telemetry-store ergonomics, MQTT, Playwright for SCADA dashboards.

LOCAL-FIRST

Your code, schematics, and probes stay on your machine.

A desktop installer. Conversations go to the LLM provider you configure — Anthropic, OpenAI, or a local model.

A real answer

"Need a low-power STM32 board for a BLE gateway."

Give InteJer your power budget and use case. It resolves the part, runs the current-draw math, and lays out real dev boards with trade-offs — then makes a call.

Low-power STM32 board for 3.3V / 100mA budget
need a low power stm32 board, my battery budget is 3.3V, 100mA — for a BLE gateway. Continuous BLE radio, moderate processing, 1–3 connections, single sensor + phone app, days (1–7) of runtime, rechargeable with frequent access. Dev board for faster prototyping, larger footprint is fine.
Calledpartfetch_resolve_partpart_number=STM32WB55 · mfr=STMicroelectronics

Board options for a BLE gateway

Budget analysis: 3.3V × 100mA = 330mW. BLE-gateway continuous mode is typically 20–40mA active + 5–15mA radio = 25–55mA. You have headroom.

Option 1: P-NUCLEO-WB55

Recommended

STM32WB55RG with integrated BLE 5.2 radio

SpecValue
CoreDual-core Cortex-M4 (64MHz) + M0+ (32MHz)
Flash / RAM1MB / 256KB
BLEIntegrated, BLE 5.2, −96dBm RX sensitivity
Power (Run)~20mA @ 64MHz
Power (Radio TX)5.8mA @ 0dBm
Sleep modes0.4µA standby, 1.7µA stop
Form factorArduino UNO headers + ST morpho
Price~$30

Why this one: Dual-core architecture lets the M0+ handle the BLE stack while the M4 runs your gateway logic. Integrated radio eliminates RF design. Best BLE 5.2 support in the STM32 lineup.

Option 2: P-NUCLEO-WB55-USB

Dongle

Same chip, compact USB dongle form factor

DifferenceValue
Form factorUSB dongle (~50mm × 20mm)
HeadersNone (test points only)
PowerUSB or external via pin headers
Price~$15

Trade-off: Best for portable testing, but harder to wire sensors. Good if your gateway is mobile/portable.

Option 3: NUCLEO-WB35RG

Cost-optimized

STM32WB35CE — smaller memory variant

SpecValue
Flash / RAM256KB / 128KB
Price~$20

Trade-off: Sufficient for the gateway, but tighter if you need OTA firmware updates or large buffers.

My recommendation: Option 1 (P-NUCLEO-WB55)

For prototyping a BLE gateway with 1–3 connections, the full Nucleo-68 gives you:

  • Arduino shield ecosystem for adding sensors / logic
  • Onboard ST-LINK for debugging
  • Enough headroom for future features
  • Best documentation / examples

Does this direction work for you, or do you want to explore any of the alternatives?

Build·GLM-5·40s
A real buildBeta

"Here's my quad spec — design the board and the frame."

InteJer read a full quadcopter design spec and produced a fabrication-ready power-distribution board — schematic, routed PCB, and 3D model — alongside a parametric, printable airframe.

  • Schematic + routed 2-layer PCB in KiCad
  • 3D board preview and rendered assembly
  • Parametric 290 mm frame + printable STLs in FreeCAD
  • Fab pack: Gerbers + CPL + BOM + STEP

Schematic, PCB & 3D-model generation is currently in beta.

power-board · 290 mm research quad
Built for the 290 mm autonomous research platform (Pi 5 companion, Pixhawk FC, DJI O4) — 66 × 36 mm with a Pixhawk-style power connector. Functionally complete: 47 components, routed, fab files generated, 0 DRC errors (15 cosmetic).
01SCHEMATICKiCad
SCHEMATIC — 290 mm research quadopen full size ↗

Full power-distribution schematic — 47 components across the power tree, regulators, and connector breakout.

02PCBKiCad
PCB — 290 mm research quadopen full size ↗

Routed 2-layer board, 66 × 36 mm, with a Pixhawk-style power connector and onboard current sensing.

033D BOARDKiCad
3D BOARD — 290 mm research quadopen full size ↗

3D board preview — component placement, courtyards, and mechanical clearances verified.

043D RENDERKiCad
3D RENDER — 290 mm research quadopen full size ↗

Rendered assembly with regulators, bulk caps, and passives populated on the board.

05FRAME CADFreeCAD
FRAME CAD — 290 mm research quadopen full size ↗

Parametric 290 mm X-frame with printable parts — arms, top/bottom plates, and camera cage.

47 components·routed·0 DRC errors·Gerbers + CPL + BOM + STEP
vs. The general-purpose AIs

Why a hardware-specific agent.

Cursor and Claude Code are excellent general-purpose coding agents. They aren't built for what runs on the silicon.

CapabilityInteJerCursor / CopilotClaude CodeChat AIs
Reads manufacturer datasheets by part number
Pulls 3D CAD (STEP / IGES) for components
Drives PlatformIO builds
Flashes & debugs via J-Link / ST-Link
Native ESP32 / STM32 / Zephyr knowledgeskillspartialpartialpartial
ThingsBoard + MongoDB telemetryMCPsmanual
Local desktop app (not cloud-only)
AI UI design (Stitch)
* privacy: code only leaves your machine to whichever LLM provider you choose
Who it's for

Built for lean teams shipping the whole IoT stack.

If you're an IoT startup that's resource-constrained on the stack — a small team that can't afford to build firmware, cloud, and frontend separately — and shipping speed is what wins, then InteJer is the best fit for you.

Small, resource-constrained teams

You can't afford to staff firmware, cloud, and frontend as three separate hires. One engineer plus InteJer covers the whole stack.

The full stack, end to end

From the hardware layer to cloud telemetry to the dashboard your customer actually sees — one agent that understands every tier, not three disconnected tools.

When shipping speed is the moat

Move from prototype to production without context-switching between toolchains. The faster you ship, the faster you learn.

One team. One agent. The whole stack — hardware to frontend.

Pricing

Start free. Upgrade when you ship.

Every plan starts with an account — sign up, download the IDE, and you're on the Free plan. Upgrade any time from your dashboard.

Free

$0/mo

Everything you need to try InteJer on a real project.

  • The full InteJer IDE
  • Bring your own model API keys
  • Limited agent usage
  • Community support
Start for free
Most popular

Pro

$20/mo

For engineers shipping with InteJer every day.

  • Access to premium models
  • Unlimited tab completions
  • Full embedded toolchain
  • Email support
Upgrade to Pro

Pro+

$40/mo

3x more usage for heavy hardware-agent workloads.

  • Everything in Pro
  • 3x agent usage
  • Higher hardware-agent limits
  • Priority support
Upgrade to Pro+

Ultra

$100/mo

Maximum throughput and early access to what's next.

  • Everything in Pro+
  • 20x agent usage
  • Early access to new features
  • Priority support
Upgrade to Ultra

All plans require an InteJer account · cancel any time

Hardware deserves better tools.

Download InteJerRead the docs