DOMAIN LABS

Sheet A-101 · Plant layout

Illustrative floor plan of a motor plant: receiving, stores, kitting, winding, an assembly line with five stations, test cells, QA lab, pack and ship, and office with the MES and ERP servers. Not a client facility.123456789101112ABCDEFG6000600060006000600060006000600060006000600066 00060006000600060006000600036 000MATERIAL FLOWOPERATORHMISCANNERLABEL PRINTER051015MNPLANT LAYOUT · ILLUSTRATIVENOT A CLIENT FACILITYDRAWN FOR DOMAIN LABSSHEET A-101SCALE 1:400RECEIVING01 · DOCKS 1–2PACK & SHIP08 · DOCKS 3–5STORES02 · RACKINGKITTING03W-01W-02W-03W-04WINDING04 · 4 WINDERSST-01ST-02ST-03ST-04ST-05ASSEMBLY LINE05 · 5 STATIONSTC-1TC-2TC-3TEST06 · 3 CELLSQA LAB07OFFICE09IT · MES/ERP
Illustrative plant layout · not a client facilityDrag across any station for a readout

We build the software your operation runs on, and set your team up to own it.

We build it with AI coding agents, connect it to your equipment and the rest of your systems, and work alongside your team until they own the codebase. That includes a production MES and ERP for an electric motor manufacturer.

See what we’ve built

Most plants already have an ERP. The work happens around it.

Schedules live on a whiteboard or in a spreadsheet. Travelers and quality records are on paper. Test results stay on the test stand. People re-key numbers from one system into the next, and one or two people know how it all fits together.

AI tools now make it easy for anyone to add another script or app on top. Each one works on its own. Without shared names, schemas and interfaces, every addition makes the whole harder to trust, and unsafe for an AI agent to touch.

So we start underneath: one naming convention for parts, lots, work orders and stations, schemas for the data that matters, one typed interface between systems, and MCP servers that let agents read by default and ask a person before they write.

Fewer than 1 in 4
large manufacturers report their MES fully connected to ERP, quality and the plant floor.Rockwell Automation survey, 2026Vendor survey, mostly large firms
60% vs 18%
of small and mid-sized manufacturers name lack of in-house expertise, against cost, as a barrier to AI.MEP network surveys, 2026Not a representative sample
Illustrative diagram. Today, schedules, travelers, quality records, test results, inventory counts, supplier email, controllers and a one-off script connect to the ERP and to each other by hand. With shared contracts, they connect through one layer of shared names, schemas and a typed interface, and an AI agent connects through an MCP server that reads by default and needs a person to approve writes. SHARED NAMES · SCHEMASONE TYPED INTERFACE MCP · READ BY DEFAULTWRITES NEED A PERSON SCHEDULEWHITEBOARD + SHEETSCHEDULEMES TRAVELERSPAPERINSTRUCTIONSSTATION SCREENS QUALITYBINDERQUALITYON THE UNIT RECORD SOMEONE’S SCRIPTON A LAPTOPSCRIPTUSES THE INTERFACE TEST STANDCSV ON A USB STICKTEST RESULTSON THE UNIT RECORD INVENTORYCOUNT SHEETSINVENTORYBY LOT SUPPLIER EMAILSOMEONE’S INBOXRECEIVINGPO AND LOT CONTROLLERSNOT CAPTUREDCONTROLLERSSIGNALS RECORDED ERP AI AGENTNO SAFE WAY INTHROUGH MCP RE-KEYEDUSB EXPORTCOPY-PASTEPAPER
Illustrative. Not any client’s systems.

The MES and ERP

We built the MES and ERP an electric motor manufacturer runs on. It covers production, quality, purchasing, inventory, and shipping. The client owns the code, database, and deployment pipeline.

Walk the plant

Built alongside
the people on the line.

Run a unit through the plant. Each screen is a simplified, illustrative version of what an operator sees at that station. Every scan and result lands on the unit’s record, so the unit can be traced back to the material lots it was built from, and each lot forward to every unit it went into.

  1. 01

    Purchasing and receiving

    Purchasing and inventory run in the same system as production. Material lots are recorded as they come in.

  2. 02

    Work orders and instructions

    Each station gets its work order and the operator instructions for that step. Serial and lot scans attach parts to the unit.

  3. 03

    Equipment and test

    We connected controller signals to the MES so tests and measurements stay attached to the production record.

  4. 04

    Quality

    Quality workflows and the audit history live on the same records as production.

  5. 05

    Shipping

    Shipping runs from the same data. The client owns the code, the database, and the deployment pipeline.

We build it so
your team can own it.

01 / Your codebase

Your team owns it

Your developers can contribute to the codebase while we build it, and take over ownership when they’re ready. You keep the code, the database, and the deployment pipeline.

02 / Custom MCP servers

Data across the organization

We build MCP servers that connect data across your organization, so an AI application can work with it. Each server sets what the application can read, what it can do, and which steps need a person’s approval.

03 / Agentic processes

Issues caught early

We set up agentic processes that check the work as it happens and catch issues early. When an agent finds something, it goes to the person who can fix it.

What we
work on.

We build custom operations software with AI coding agents, connect it to the systems and equipment around it, and set the client’s team up to keep building it.

Domain Labs is run by Matthew Rundle. Before Domain Labs, Matthew built and led a team of six data scientists at Expedia / Vrbo.

  • MES and ERP in production: work orders, operator instructions, serial and lot traceability, quality, and machine interfaces.

  • Agents that check the work as it happens and send what they find to the person who can fix it.

  • The data an AI application can read, the actions it can take, and the steps that need a person’s approval.

  • Scribe: draft notes and proposed billing codes that the clinician edits and approves.

  • Custom applications for businesses with a lot of moving parts.

Work with us
Let’s talk about your operation

What do you
need to build?

Tell us what runs the operation today and what you want to change. Or write to us directly.

contact@domainlabs.dev
What runs the operation today? optional