Anthropic has released Claude for Manufacturing—a plugin for manufacturing companies with 17 skills and four connectors to factory systems. Imagine Semen, an engineer at the fictional “Pigsty No. 42” lab: he receives a customer inquiry, but before preparing a quote, he needs to check costs, production data, and material availability. The plugin can prepare the necessary materials, but leaves the process’s completion to a person. Oink, that boundary matters more here than an impressive list of capabilities.
Anthropic’s release came out on 12 November 2026. The plugin brings together a set of Claude skills, connections to enterprise data sources, and a human review step. So this isn’t one factory feature, but several ready-made workflows that work with a company’s specific data. Anthropic also says organizations can adapt the skills to their own workflows.
What tasks do the 17 skills cover
The skills cover sales, quality, operations, procurement, design, and configuration. Some perform a single operation; others gather information in sequence and prepare an output for the next human decision. For example, to calculate a new order, Claude reads the customer inquiry, identifies missing information, compares costs with similar orders, and prepares a quote.
For Semen, this means an inquiry doesn’t have to turn straight into numbers that nobody has checked. Claude can flag what’s missing and put together a draft based on available data; the final decision on the quote remains with the employee. So the skill helps get through the preparatory steps, but doesn’t remove a person’s responsibility for approval.
Another workflow deals with a failed quality inspection. Claude can document the defect, look for similar cases in previous reports, organize possible causes, and prepare a draft corrective action report. In the event of equipment failure, it analyzes supplied sensor exports, repair records, photos, and the manual, then ranks likely causes for a technician to investigate. This is preparation for diagnosis, not confirmation of the cause in place of a specialist.
In procurement, the plugin can draft a request for quotations when changing suppliers and compare the proposals received. In Autodesk Fusion, it generates options based on the existing design, evaluates them against the engineer’s goals, and prepares a decision memo. Another scenario compiles a production report: output compared with plan, equipment downtime, and its cost. These are different work tasks, brought together by a common principle—Claude prepares an analysis or document, and a person checks the result.
Where Claude gets manufacturing data
The plugin has connectors to Autodesk Fusion, Infor, Microsoft 365, and Oracle NetSuite. Through them, Claude reads information from systems the company already uses. In Microsoft 365, it can search for procedures, supplier emails, and reports in shared folders and email. For Oracle NetSuite, the description lists orders, inventory, production stages, and cost of goods—data needed, among other things, to calculate a quote.
In Infor, Claude can check purchase orders and inventory, while Autodesk Fusion is used to work with design options. If a manufacturing system doesn’t have a connector, Anthropic describes a workaround: regularly export data to files and store them in Microsoft 365. Claude then works with those files. This is an important distinction: data being available through files doesn’t mean the plugin has a direct connection to every system in the enterprise.
Semen’s imaginary technical pigsty shows well why the list of connectors matters in practice. To prepare a response about an order, Claude has to rely on sources that are connected or supplied as files. If the necessary information isn’t there, the system can’t magically fill the gap; in the order calculation scenario, it can point out missing data. That’s why a ready-made workflow depends not only on the skill, but also on access to relevant documents.
Reading data and making changes are different permissions
The Infor connector included in Claude for Manufacturing is intended for reading records. A separate plugin released by Infor has broader capabilities: Claude can make changes using the user’s permissions in Infor. For example, in Anthropic’s scenario, a steel shipment is delayed by two weeks. The procurement specialist asks which customer orders will be affected and how much steel is available at other sites.
After approval from the procurement specialist, Claude can place an order with another supplier and reschedule production in Infor. However, this example applies specifically to the separate Infor plugin, not the read-only connector included in the manufacturing plugin. Access is limited by the procurement specialist’s own permissions in the system. For deployment, this isn’t a minor footnote: the ability to read a record and the ability to change it are different levels of permission.
A person completes the process; the equipment stays out of it
Anthropic describes the skills as stopping before a decision that a person needs to review. The estimator issues the quote themselves, and the quality engineer confirms the correction. For verification, Claude links figures and claims to the source data and documents. This makes it possible to check the prepared material against its sources, but doesn’t turn a draft into an automatically approved result.
There’s a separate boundary, too: Claude for Manufacturing doesn’t control equipment. For programmable machines, Anthropic is developing the Model Hardware Standard—a research standard intended to describe a machine tool’s measurements, settings, and safety limits in a consistent format. This isn’t a control feature added to the plugin, but a separate line of work at the company.
For Semen and his imaginary team, this distinction suggests what to expect: the plugin helps prepare an analysis, compare available records, and put together a document, but it doesn’t replace a specialist’s review or command the machine tool. Anthropic cites a Manufacturing Leadership Council survey from 2025: 79% of respondents named manually entered data among their sources of manufacturing data. The company presents its skills as a starting point and promises to add new manufacturing scenarios.
Installation and availability
Claude for Manufacturing is available on the Team and Enterprise plans. To install it, the user tells Claude, “Set us up”; Claude then asks which systems the team uses, checks the connectors, and reports which skills won’t yet be able to work fully. So setup starts with a list of systems and a check of available connections—not with the assumption that every scenario is ready to go right away.
Anthropic announced a technical webinar with a demo of the plugin on 17 November at 11:00 US Eastern Time, or 19:00 Moscow time. The announcement doesn’t specify a separate price for the plugin. For a manufacturing team, the main practical question remains concrete: which of its data sources are connected, which processes the skills cover, and exactly where a person needs to check the result before it becomes an action.









