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)
ReleaseAPI / Product · ActionableThe Decoder→ original

Anthropic Gives Claude Access to Factory Software Through 4 Connectors and 17 Skills

Anthropic has released Claude for Manufacturing, a plugin that brings AI into factory software so that data no longer has to be copied by hand. The four connectors are ready-made bridges to Autodesk Fusion, NetSuite, Microsoft 365 and Infor, while the 17 skills are prebuilt workflows, such as pricing a quote from actual warehouse stock. Results are tied to source documents, and a person reviews them before anything goes to a customer, a practical setup for industry agents.

AI-processed from The Decoder; edited by Hamidun News
Anthropic Gives Claude Access to Factory Software Through 4 Connectors and 17 Skills
Source: The Decoder. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceThe Decoder ↗
Editor-in-Chief
Key metric #1Ordering from a backup supplier and shifting production happen only after human approval
Key metric #2Available to Team and Enterprise customers; the plugin does not yet control machine tools
Key metric #3Model Hardware Standard is meant to connect agents to robotic arms and microscopes

A factory AI assistant usually runs into one problem: in chat it answers nicely, but it cannot see warehouse stock, drawings or orders. People copy data by hand from one program to another, and all the time saved goes to moving files around. Anthropic has released Claude for Manufacturing, a plugin that gives the model direct access to a factory's working software. Access runs through four universal connectors, that is, ready-made bridges to Autodesk Fusion (design), Oracle NetSuite (accounting and enterprise resources), Microsoft 365 (documents and email) and Infor (the production system). On top of the bridges sit 17 skills, preconfigured workflows: for example, assembling a commercial proposal based on actual warehouse stock or comparing suppliers.

What exactly was connected

If a needed system is not among the four connectors, its data can be passed in through file export to Microsoft 365. This is not as deep an integration, but it makes it possible to connect even what has no ready-made bridge.

What the 17 skills include, according to Anthropic's description. Claude can:

  • draft commercial proposals based on actual warehouse stock rather than "approximate" figures;
  • compare suppliers;
  • prepare quality reports;
  • suggest design options;
  • analyze sensor data and technical manuals when equipment sits idle.

The last item is the most vivid. Suppose a machine stops. Previously an engineer searched the manual for the right page, checked sensor readings and only then worked out where to look for the fault. Now the model does the initial work, and the person gets a short analysis with links to sources.

Where the AI decides and where it only proposes

The most sensitive area is orders and scheduling. Through an additional plugin for Infor, Claude can order parts from a backup supplier if the main delivery is delayed, and rebuild the production schedule. But it does this only after a person approves. All results come with links to source documents: the employee sees which data the model based its conclusion on and checks it before the document goes to a customer or supplier.

This setup is sensible. A mistake in a chat reply costs a few minutes, while a mistake in a component order or a shifted production queue can be expensive. Mandatory confirmation and source traceability turn the AI into a fast preparer of decisions, not an autonomous buyer.

What about physical machines

An important caveat: Claude for Manufacturing does not control physical equipment. The plugin is available to Team and Enterprise customers and works at the level of software and documents. For machine tools and laboratory instruments, Anthropic is separately developing Model Hardware Standard, a research preview project. It is meant to connect AI agents to devices such as robotic manipulators and microscopes through standardized drivers.

The point of the standard is to save on connection effort. According to The Decoder, it cuts setup from many weeks or months to hours or minutes. Today each new instrument requires a separate integration, and that integration is what slows adoption. Unified drivers would let devices be connected roughly like USB peripherals. At the same time, Anthropic openly acknowledges that human supervision remains mandatory, because the model's understanding of physical processes is limited.

What this means for practitioners

For developers and product managers building agents for industry, the news carries three takeaways. First, large labs are moving into vertical solutions: not a "universal chat" but a set of skills tailored to a specific sector and specific systems. Second, the pattern "the agent prepares, a human confirms, sources are attached" is becoming the standard for tasks involving real money and deliveries. Third, standardized hardware drivers may become the next point of competition: whoever first simplifies connecting instruments will win the factories.

The limits are also clear. The publication contains no data on how accurate the proposals and reports are at real enterprises, and no figures on time savings for the plugin itself. There are only four connectors, and older factory systems often live outside that list, in which case work has to go through file export. For now this is more a market claim than a proven result, and it is worth evaluating in a small pilot with your own data.

Story

Key Questions Answered

What exactly was connected?

If a needed system is not among the four connectors, its data can be passed in through file export to Microsoft 365. This is not as deep an integration, but it makes it possible to connect even what has no ready-made bridge. What the 17 skills include, according to Anthropic's description.

Where the AI decides and where it only proposes?

The most sensitive area is orders and scheduling. Through an additional plugin for Infor, Claude can order parts from a backup supplier if the main delivery is delayed, and rebuild the production schedule. But it does this only after a person approves. All results come with links to source documents: the employee sees which data the model based its conclusion on and checks it before the document goes to a customer or supplier. This setup is sensible.

What about physical machines?

An important caveat: Claude for Manufacturing does not control physical equipment. The plugin is available to Team and Enterprise customers and works at the level of software and documents. For machine tools and laboratory instruments, Anthropic is separately developing Model Hardware Standard, a research preview project. It is meant to connect AI agents to devices such as robotic manipulators and microscopes through standardized drivers. The point of the standard is to save on connection effort.

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…