Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)

Anthropic Releases Claude for Manufacturing With 17 Skills: It Estimates Orders and Hunts for Machine Faults

On November 12, Anthropic released Claude for Manufacturing, a plugin for factory calculations and fault diagnosis. It includes 17 skills (ready-made workflows) and connectors to order-tracking ERP systems such as Infor and NetSuite. When a client sends a drawing and asks for a price, Claude builds a 3D model, estimates setup time, and compares similar past orders. A pilot makes the most sense where the data already sits in the ERP.

AI-processed from TG: @disruptors_official (Дизраптор); edited by Hamidun News
Anthropic Releases Claude for Manufacturing With 17 Skills: It Estimates Orders and Hunts for Machine Faults
Source: TG: @disruptors_official (Дизраптор). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTG: @disruptors_official (Дизраптор) ↗
Editor-in-Chief
Key metric #1When a machine stops, the AI ranks likely causes by probability, and a human carries out the repair.
Key metric #2The plugin does not control hardware; robotics falls under the MHS standard.
Key metric #3The output for a drawing is a draft quote and a list of missing data.

Picture a factory whose client has sent a part drawing and asked, "How much will this cost?" To give a price, a process engineer has to work through the drawing, sketch out the operations, estimate machine setup time, and recall what similar orders cost before. That is hours of work, and such requests can come in by the dozen each day.

Claude is now taking over this paperwork side of factory life. On November 12, Anthropic released Claude for Manufacturing, a plugin for manufacturing companies. It includes 17 skills, meaning 17 ready-made workflows such as order costing or defect analysis.

It also comes with prebuilt connections to the software factories already run on: the ERP systems (which hold orders, purchasing, and inventory) Infor and NetSuite, the Autodesk engineering software, and the Microsoft office suite.

What's inside the plugin

Anthropic is not asking factories to move to a new platform. The plugin connects the model to what the plant already runs: order management, drawings, and office documents.

For practitioners, this detail matters. The main obstacle in enterprise AI deployments is usually not model quality but access to data and tools. If an agent already has connectors to the ERP and CAD (computer-aided design) systems, a pilot can start with a specific task rather than a half-year integration project.

How it works in practice

The first workflow is order costing. Claude picks the tools it needs, builds a 3D model from the drawing, breaks production into operations, estimates setup and machining time, and checks what similar orders cost before. The output is a working draft of a commercial proposal and a list of what the client forgot to specify.

The second workflow is a machine stoppage. The models take in sensor logs, repair history, photos, and the operating manual. The specialist gets a list of possible causes sorted by probability: what to check first. Claude will not fix the machine, but it can cut the time spent hunting for the fault.

Following the same pattern, the plugin can analyze defects, compare suppliers, and compile a morning downtime report. The general principle is simple: if a factory has data, a skill can be built on top of it.

The MHS link and honest limitations

In August, Anthropic introduced MHS, a standard similar to MCP but for physical equipment. Through it, Claude can control robotic arms, microscopes, and laboratory setups. The new plugin does not touch the equipment: it handles planning, calculations, and analysis, not machine control.

That makes for two products with different roles. Claude for Manufacturing thinks at the level of the office and the engineering department, while MHS could potentially give it hands on the shop floor. Together they form a picture of production automation, but for now they are two separate solutions.

The caveats are these. First, according to the author of the review, there are still few genuinely working integrations. Second, every scenario described keeps a human in the loop who approves the result: a process engineer checks the estimate, and a maintenance technician selects the likely causes of a fault. For manufacturing, that is a sensible setup, since an error in a calculation or a diagnosis is costly. Third, the description is based on a retelling, and there is no independent data yet on the accuracy of cost estimates or diagnostics. Before trusting the figures in a draft proposal, they should be checked against archived orders.

What to do with this

For developers and product managers who work with industrial clients, the takeaway is practical:

  • The market is moving toward industry-specific skill sets and connectors, not general-purpose chatbots.
  • Tasks with high volumes of repetitive analysis and a solid data history pay off fastest: order estimation, defect analysis, reporting.
  • A pilot is best built with mandatory human sign-off, and it should measure how far the AI drafts diverge from actual calculations.

Factories are not running without people yet, but part of the work of "the shop supervisor buried in paperwork" is already becoming a job for a model.

Story

Key Questions Answered

What's inside the plugin?

Anthropic is not asking factories to move to a new platform. The plugin connects the model to what the plant already runs: order management, drawings, and office documents. For practitioners, this detail matters. The main obstacle in enterprise AI deployments is usually not model quality but access to data and tools. If an agent already has connectors to the ERP and CAD (computer-aided design) systems, a pilot can start with a specific task rather than a half-year integration project.

How it works in practice?

The first workflow is order costing. Claude picks the tools it needs, builds a 3D model from the drawing, breaks production into operations, estimates setup and machining time, and checks what similar orders cost before. The output is a working draft of a commercial proposal and a list of what the client forgot to specify. The second workflow is a machine stoppage. The models take in sensor logs, repair history, photos, and the operating manual.

What are the key details on The MHS link and honest limitations?

In August, Anthropic introduced MHS, a standard similar to MCP but for physical equipment. Through it, Claude can control robotic arms, microscopes, and laboratory setups. The new plugin does not touch the equipment: it handles planning, calculations, and analysis, not machine control. That makes for two products with different roles. Claude for Manufacturing thinks at the level of the office and the engineering department, while MHS could potentially give it hands on the shop floor.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Hands-On AI Engineering · Hamidun Academy

Build & Ship Production AI Agents Without Boilerplate Code

Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.

Start 72h Free Trial →Vibe-Coding Intensive
What do you think?
Loading comments…