For hardware engineers

Your system architecture,
grounded in real parts.

Describe your system in plain English. RoboSpec lays out the architecture, then grounds every component in real distributor and manufacturer data — with calibrated confidence and full source provenance.

Try the live app
robospec — architecture + drill-in
$ pick & place, 5 kg cartons, automotive line
Controller / PLC
Siemens S7-1500
Robot arm
FANUC M-10iD/12
End-of-arm tooling
Schunk PGN-plus 80
Vision sensor
Keyence IV3-G500MA
Safety system
SICK microScan3
Conveyor
Dorner 2200 series
— hard wired  ·  - - signal / vision  ·  click a node to drill in
 Manufacturer-verified  ·   Distributor-verified  ·   AI-suggested
Vision sensor — recommendation
Keyence IV3-G500MA High confidence

Compact vision sensor for part presence and orientation before pick. Matches the required resolution and monochrome imaging for this cell.

Price est.
~$830 (distributor list price)
Lead time
In stock at a distributor
Source: Distributor-verified · Nexar — IV3-G500MA (Active, in stock)
⚠ Confidence capped where a spec can't be verified against real data — never inflated.
8
Component categories
2.4M+
Parts coverage
3
Confidence levels, from the data
~30s
To a full system architecture
How it works

Architecture first.
Real parts on click.

01 — DESCRIBE

Describe the system

One plain-English box — what it does, payload, throughput, environment. No forms to fill.

02 — REVIEW PARTS

Check the component list

Get a suggested part list first. Uncheck anything you don't need and adjust quantities — with live power-distribution and communication guidance updating as you go. Then generate the architecture.

03 — ARCHITECTURE

See the whole system

A clickable diagram in seconds — every component as a node, wired with real connection types: EtherNet/IP, safety bus, IO-Link.

04 — DRILL IN

Verify & export

Click any node for its reasoning, confidence, and sourcing — manufacturer-verified, distributor-verified, or AI-suggested. Export the BOM as .csv.

What's generated

Credible output,
not plausible guesses.

OUTPUT 01

System architecture diagram

Your whole cell as connected nodes — controller, robot, tooling, vision, safety, conveyor, sensors, power. Hover a wire to see its protocol.

OUTPUT 02

Real, sourceable parts CORE

Every recommendation is grounded in live distributor and manufacturer data — never an invented model number. Your pick is never silently swapped: if it can't be verified it's labeled AI-suggested, with up to two real alternatives to choose from.

OUTPUT 03

Calibrated confidence

HIGH / MEDIUM / LOW, derived from the data itself. A missing spec caps confidence — it's never inflated.

OUTPUT 04

Source provenance MOAT

Every part is tagged manufacturer-verified, distributor-verified, or AI-suggested — with the requirement it was matched against. You always know where a number came from.

OUTPUT 05

Watch-outs, price & lead time

Per-part price and lead time, risk flags surfaced up front, plus a rough total system cost range for the whole cell.

OUTPUT 06

Full BOM export (.csv)

Export one component or the whole system — brand, part number, confidence, price, and a lookup link for every line.

Why the confidence label is trustworthy

Scored by the data,
not the model.

HIGH

Part is active, in stock, and every spec you gave is confirmed against real distributor or manufacturer data. Safe to source.

MEDIUM

A real part, but one or more specs couldn't be fully verified, or it's obsolete or out of stock. Capped here on purpose.

LOW

No verified match — the suggestion is AI-inferred and flagged as a starting point to confirm before sourcing. Honest by default.

A deterministic validation layer scores each part against your specs and live distributor & manufacturer data — the confidence you see is derived from that check, not the model's opinion of its own answer.

Why not just ask ChatGPT?

Generic AI guesses.
RoboSpec verifies.

Capability
RoboSpec
Generic AI
Distributor search
Manual speccing
Start from system requirements
Real, in-stock part numbers
Grounded in a live catalog
Confidence calibrated to data
Flags obsolete / out-of-stock
Source provenance (verified vs inferred)
System-level architecture & connections
Full BOM export

The moat isn't generating a spec — it's output you can trust. Real part numbers, honest confidence, a clear record of where every recommendation came from, and no incentive to favor whoever pays to be listed.

Who it's for

Built for people who
live in the field.

Systems integrators

Turn a client brief into a full architecture and a real, sourceable BOM in minutes — with confidence flags that tell you exactly what still needs a phone call.

Automation engineers

Scope new cells fast and reach procurement with verified, in-stock part numbers instead of a rough estimate a generic AI tool may have made up.

Plant & OEM managers

Get a structured spec with provenance on every line — so you know which numbers are catalog-verified and which are still a starting point.

Any hardware developer

Describe what you're building and let the model reason out the architecture — then grounded part data cuts the hours you'd otherwise spend searching for each component yourself.

Where we're headed

From a trusted BOM to
the AI sandbox for hardware.

Today, RoboSpec owns the spec-and-decision layer — the trusted parts and BOM at the foundation of a build. From there it expands outward: the AI sandbox engineers live in across the whole cycle, and — in the process — the un-scrapable dataset of how hardware actually gets designed.

NEXT

Own the full build ROADMAP

Expand from spec and BOM across the whole cycle — architecture, CAD layout, simulation, procurement — by orchestrating the tools engineers already use (SolidWorks, PLM, simulators), not rebuilding them. One sandbox, end to end.

NEXT

Cross-brand compatibility ROADMAP

Normalize parts onto ECLASS / ETIM and read interoperability files (EDS / ESI / GSD, OPC UA) to answer "what's the Allen-Bradley equivalent?" and "will these actually talk to each other?" — real evidence, not catalog adjacency.

MOAT

A decision dataset that compounds ROADMAP

Every Keep / Swap / Drop sharpens the next recommendation — a proprietary record of how engineers actually choose parts, so accuracy and confidence improve the more the tool is used.

NORTH STAR

The physical-AI data layer ROADMAP

Over time that record becomes the scarce, un-scrapable corpus of how physical hardware gets designed — the foundation for AI that can reason about the real world, not just code.

The AI sandbox
for hardware is here.

Try the live beta, or join the list for early access.

Try the live app