First principles for embedded and edge software

Firmware architecture that's structured, observable, and AI-ready — built in weeks, not months.

We spent 25 years watching the same problems repeat. We decided to solve them structurally. EmbedIQ is the open-source architecture layer for embedded and edge systems. Every firmware layer is a clear contract — that's why humans and AI agents can build together in weeks, not months. RTOS, bare-metal, or Linux — one architecture.

Run firmware on your laptop Thermostat + industrial gateway — both running on Linux today. No board required.
AI observability built in EU AI Act Art.12/13 logging. Four AI event constants. Tamper-evident .iqtrace audit log.
Safety-capable by design ISO 26262, IEC 62304, IEC 62443, AIS-190. safety_class on every FB. CycloneDX SBOM.
What exists today: Phase 1 + Library Architecture complete The framework runs. All tests green on main. Thermostat (5 FBs, thermal FSM) and industrial gateway (6 FBs, edge-to-cloud) both running on Linux — clone the repo, 3 commands, no hardware required. Phase 2 active: Service FBs and tooling. Phase 3 next: FreeRTOS, ESP32, fb_cloud_mqtt, fb_ota on real hardware.
What we want developers to feel

Debugging firmware without printf

Run the same logic on a real device and on your laptop. Watch the system move. Catch the bug before it becomes folklore.

$ ./build/thermostat_host
[ 1.001] boot phase 1 — fb_timer RUNNING
[ 1.002] boot phase 2 — fb_watchdog RUNNING
[ 2.001] TX fb_sensor → fb_ctrl MSG_TEMP_READING seq=3
[ 2.002] FSM fb_ctrl.temp_monitor NORMAL → WARNING seq=4
[ 2.003] TX fb_ctrl → fb_cloud MSG_CLOUD_PUBLISH seq=5
[ 2.004] OBS queue_depth=2 source=fb_ctrl
Less "why is the watchdog angry again?"
More "ah, the message order changed at 87°C — found it."

Build connected devices in weeks, not months.

The problem

The substrate is not the problem.
Everything above it is.

Teams keep rebuilding the same infrastructure above the substrate — on every RTOS project and on every Linux gateway. The problem is structural, not skill-based.

🧵

Task spaghetti

Threads, callbacks, flags, and globals slowly grow into firmware that only its original author can safely touch.

🪛

Same plumbing, every project

OTA, telemetry, watchdog, NVM, cloud glue — rebuilt from scratch instead of reused as clean building blocks.

🐛

Debugging by archaeology

Too much embedded debugging depends on printf, timing luck, and whoever remembers the original intent.

🤖

AI generates code, not architecture

Without strong structure, AI tools just help teams produce inconsistent firmware faster. The structure has to come first.

Why now

This is the right moment to rethink embedded software.

  • → Mbed OS is archived July 2026. Thousands of developers need a new architecture home — now.
  • → AI development is here. But embedded systems lack the structure agents need to generate reliable firmware.
  • → RTOS projects stop at the kernel layer deliberately. The application architecture above them has always been left to each team.
  • → The closest alternative (QP/C) has been downloaded 60,000 times a year for 19 years — and uses GPLv3. Developers want this. The license blocks them.
The gap has always been real. The timing just got urgent.

RTOS gives you primitives.
Architecture gives you leverage.

EmbedIQ is being built to make embedded firmware structured, observable, and AI-codeable — from the architecture level up, not as a plugin or afterthought.
How we think about it

A simple stack, built from first principles.

Hardware and RTOS stay where they belong. EmbedIQ is the structured application layer above them.

Commercial toolsStudio · Cloud · AI Coder — future commercial layer
Reusable componentsfb_cloud · fb_ota · fb_telemetry · fb_provisioning
EmbedIQ Core ✓ engine runningFunctional Blocks · message bus · FSM engine · Observatory · testability
OSAL / HALFreeRTOS · Linux POSIX · Zephyr · RISC-V — same application code, any target
SubstrateFreeRTOS · Zephyr · bare-metal

See the full architecture, FB topology, and compliance map →

🧠 AI-first by architecture

Typed, machine-readable components enable AI to generate working firmware from natural language. Learn how →

👁 Zero-instrumentation observability

Every event captured automatically. Studio visualizes it. AI analyzes it. Learn how →

💻 Host simulation first

Same code runs on laptop or device. Full test suite in CI under 5 seconds. Learn how →

🔒 Contracts frozen at v1

No breaking changes. Write a component today, it compiles in 5 years. Learn how →

Every layer. One model.

Each tier, you either reuse an open-source block or an AI coding agent writes it — conforming to a typed contract.

🖥️ Hardware + BSP

✓ Reuse board support — or — 🤖 AI writes it · typed contract

🌉 OSAL / Abstraction

✓ Reuse abstraction layer — or — 🤖 AI writes it · typed contract

🔌 Drivers + Services

✓ Reuse (MQTT, OTA, telemetry) — or — 🤖 AI writes it · typed contract

📱 Application

✓ Reuse blocks + your logic — or — 🤖 AI writes it · typed contract

Every layer is a clear contract. That's why a human + an AI coding agent build it in weeks, not months.

Explore architecture → EmbedIQ | Studio EmbedIQ | Cloud
🛡️

Compliance-capable by architecture

Every FB carries a safety_class field. The Observatory produces a tamper-evident .iqtrace audit log automatically — no developer instrumentation needed. CycloneDX SBOM, SPDX 2.3, EU AI Act logging.

ISO 26262 IEC 62304 IEC 62443 EU AI Act AIS-190

Full compliance map →

What's ready now

EmbedIQ

The open-source architecture layer for embedded and edge systems.

Explore →

EmbedIQ | Cloud

Fleet operations, device management, and OTA updates. Invite-only early access.

Request Access →

EmbedIQ | Studio

Device debugging, trace analysis, and AI-assisted root cause. Invite-only early access.

Request Access →
Get involved

The source is open.
Read the code. Run the tests.

Everything is on GitHub. Clone the repo, run the examples (Linux today, expanding to other targets), read the architecture docs, and see for yourself. No signup needed.

⭐ Star the repo It tells us you care about the problem — and helps shape which targets and Driver FBs we prioritize next.
💬 Tell us what to build Which hardware target do you need? Which Driver FB is missing? GitHub Discussions is where priorities get set.
🐛 File a real bug The engine runs on POSIX. If something breaks, open an issue. Real behavior beats design theory every time.
👁 Watch the build ESP32 and Pi 4 targets are next. Watch → All Activity to follow the commits as Phase 2 ships.
🔀 Contribute FreeRTOS OSAL, ESP32 HAL, Driver FBs — Phase 2 needs engineers who have shipped production RTOS firmware. Read CONTRIBUTING.md.
View the code on GitHub → Apache 2.0 · No account needed to read · GitHub account needed to comment or star
Expert review done. Architecture locked. What counts now is real hardware, real drivers, real bugs. That is what Phase 2 is for.