Sign up
Menu
An orange robotic fox on a workbench beside its CAD model on a monitor

YOUR AI HARDWARE ENGINEER

Build ideas.
Make real hardware.

From an idea in chat to a prototype in your hands.
Design the electronics and enclosure, create build files, and prepare your device for prototyping - all through conversation.

Enter Your Workshop
YOUR IDEA. YOUR NEXT INVENTION.See how it works

FROM IDEA TO PROTOTYPE

Turn your idea
into something real.

You don’t need a blueprint to begin.
Start with the problem your device should solve.

01 /

Let's chat about your idea

What should your device do? Who is it for? Discuss features, dimensions, and constraints with AI.

02 /

Design through conversation

Choose components, refine your schematic, and adjust the enclosure using everyday language.

03 /

Bring it all together

Keep schematics, models, and documents alongside your chat. Review your files and prepare to build a prototype.

LESS SWITCHING. MORE CREATING.

Meet your AI
hardware engineer.

The conversation on the left. Your design on the right.
Everything your device needs,
in one workspace.

Projects Air quality monitorFoxy Project
Project chat
I want to design a small fox-shaped desk gadget with a sweeping tail. It should light up when someone walks by and charge via USB-C.
FoxyCAD

I’ll start with the fox’s shape, then fit the electronics inside it. The tail gives us room for the battery, while the body can hold the controller and motion sensor.

Could the eyes light up? And can the shell open so I can replace parts?
FoxyCAD

Yes. I’ve added two LEDs behind the eyes and split the enclosure into removable sections. Here’s the proposed wiring and an initial parts list:

Explore what you can create
Exploded view of an orange robotic fox showing its enclosure and electronic componentsAIR / 01 VISUAL CONCEPTAn idea takes shape
Your work, in familiar formatsKiCadSTEPSTLPDFMarkdown

ENGINEERING PRACTICES

Built on a clear process.

From measurable requirements to prototype reviews, every project receives engineering guidance, reference documents and a consistent file structure.

Explore the process and sources

GOOD QUESTIONS. CLEAR ANSWERS.

Frequently asked questions.

Which parts of hardware development does Foxy actually do?

Foxy works on requirements, architecture, electronics, enclosure, firmware and prototype documentation within one project. Every project receives stage instructions and reference guides. Managed KiCad DRC checks are available when the service is configured; other operations depend on the tools in the project environment. Missing tools and unfinished checks must stay visible.

How autonomous is it, and where is a person still involved?

The agent can create and revise files, use available tools and continue from saved state. You make consequential choices and provide hardware observations; independent reviewers and manufacturers have their own roles. Chat and file logs show parts of this work, but Foxy does not yet measure every human intervention or promise an unattended result.

Can another engineer reproduce or continue my project?

Projects keep editable sources, stable part IDs, revision records and check evidence. The workflow calls for build instructions, tool versions and release manifests. Files organizes electronics, mechanics, firmware, references, documentation and manufacturing packages. Existing project paths are preserved. This supports handoff, but is not a complete automatic replay system.

Can Foxy handle complex, high-speed, RF or safety-critical boards?

There is no validated universal complexity limit or published benchmark covering those classes. The workflow requires project-specific constraints and specialist checks. A successful simple-board example does not establish performance for high-speed, RF, precision analog or safety-critical designs.

What proves that a design is good — beyond powering on?

The process asks for measurable requirements, calculations, review findings and evidence tied to the current design revision. Foxy checks registered files and report hashes for consistency; it does not independently prove that every report is correct. Clean DRC or a powered-on board does not prove electrical performance, reliability or certification. Physical claims need actual measurements.

Does it check signal integrity, EMI, temperature and enclosure fit?

These are covered in the design guidance where applicable, with assumptions and required checks recorded. DRC does not validate all of these effects. Specialist simulation, dimension checks and physical testing require the corresponding tools and evidence. Foxy does not currently provide a universal managed SI/PI/EMI/thermal verification pipeline.

How are component choices, pinouts and availability checked?

The agent is instructed to record exact part numbers, packages, datasheet revisions and source sections, and to review pinouts and operating margins. BOM prices retain their source or remain unknown. Foxy does not yet have an independent verified component database or universal live-stock service; substitutions need a fresh impact review.

Does the AI test its ideas, or just write a convincing report?

It can run the managed KiCad DRC service and other available project tools. Instructions require actual results, input revisions and separate evidence, and distinguish calculations, agent review, tool results and measurements. A managed SPICE loop and independent schematic review service are not currently included. A written PASS alone is not independent proof.

How much will the complete project cost?

Foxy shows available token-usage estimates and sourced BOM prices, including unknown prices. These are not an all-in quote: manufacturing, shipping, taxes, external services, testing and engineering time may add costs. There is no guaranteed project budget cap or full-project cost prediction. You can stop the agent and keep the files already created.

How is this different from using a regular AI chat?

Foxy brings project files, built-in viewers, managed DRC, BOM, shared access, saved context and file history into one workspace. Its project instructions now include stage gates, reference guides and consistent output locations. We have not established a measured speed or quality advantage over a regular chat with equivalent tools.

How repeatable are checks, and can the AI miss errors?

Registered checks are tied to file hashes so changed inputs can invalidate old results. Repeatable file-consistency checks do not measure the accuracy of AI review. Foxy has no published false-positive or missed-error benchmark for that review. Ask for the specific finding, affected file, evidence and remaining action; independent review remains valuable.

Can I upload confidential schematics or restricted PDKs?

Foxy provides authenticated access and project roles, but AI requests can involve external model processing. This is not a promise of offline operation, no-training terms or compatibility with your NDA or PDK license. Confirm the deployment, provider terms and your organization’s data requirements before uploading restricted material.

Will it lose context or break another part of a large project?

Saved project state, decisions, part IDs and interface records guide new and resumed tasks. The process requires an impact review when requirements, components or interfaces change. This reduces reliance on chat history, but incomplete dependency records and model mistakes are still possible. Older file paths are preserved rather than silently migrated.

Does this replace an engineer, and can beginners use it?

Beginners can start with an idea and use the documents to understand decisions. Engineers can direct the work, inspect evidence and collaborate on the same project. Foxy supports engineering work; it does not guarantee expert judgment, job replacement or reduced review time. A dedicated teaching mode is not currently included.

Can I send the output straight to a factory?

Use the prototype-package review first: reconcile sources, exports, BOM, placement, firmware and assembly instructions, and resolve missing or failed checks. A prototype-order candidate describes recorded digital readiness, not a production-qualified design. Supplier acceptance, ordering, assembly and physical tests are separate steps. Foxy does not automatically place manufacturing orders.

Which best practices does the AI hardware engineer use?

Every project receives an evidence-led process and discipline guides informed by NASA systems engineering, Altium and Zonenberg PCB reviews, manufacturer guidance, KiCad library conventions, OSHWA documentation practices and Instrumental’s engineering-build guide. Explore the process, source links and descriptions. The agent is instructed to read the relevant guide on each new or resumed task.

HAVE AN IDEA IN MIND?

Your next invention
starts with a conversation.

Open FoxyCAD