Anthropic brings Claude to meetings and local data centers
Anthropic opened a private beta for self-hosted agents and launched a manufacturing plugin and Google Meet integration. What it means for data control.
On November 12, Anthropic released a set of updates that move its artificial intelligence models deeper into daily business operations. The company opened a private beta for self-hosted Claude Managed Agents, launched a tool that joins video calls, and introduced a targeted plugin for the manufacturing sector. For business operators, the announcements shift the focus from what a model can write to what an agent can execute securely inside a company network.
Agents inside the company perimeter
Until now, companies using Anthropic to run automated agents had to rely on the provider to handle credentials and session records. The new self-hosted Claude Managed Agents option changes this setup. According to Help Net Security (opens in a new tab), customers can now run the agent execution harnesses, sandboxes, credentials, and session history inside their own cloud accounts or on-premise data centers.
This architecture centers on a customer-operated server and requires a PostgreSQL 16 or later database, as noted by Unite.AI (opens in a new tab). The setup isolates sensitive credentials from the sandbox where the agent code actually runs. Anthropic built the system on the premise that human oversight of every automated step is impossible, so "the limits have to be built into the system itself." Customers can keep their credentials in their own secrets manager, and a proxy outside the sandbox attaches them to approved requests.
Self-hosting the service does not mean the model runs locally. Model inference requests still leave the network. Customers can route these requests to Anthropic directly, Amazon Bedrock, Claude Models on Google Cloud, or Microsoft Foundry. The prompts and tool results go to the chosen provider, while the memory and credentials stay behind the customer firewall. To tighten security further, each environment can be locked to a specific list of allowed hosts, blocking requests to any other destination. The system also exports logs over OpenTelemetry for internal security monitoring.
Active participation in meetings
Anthropic also expanded its tools for knowledge workers. The company launched Claude Meetings, a feature that allows the model to join Google Meet calls as an active participant. Computerwoche (opens in a new tab) reports that Claude can take notes, answer questions in the chat window, retrieve context from connected applications, and create project schedules.
The meeting notes and action items appear in the browser version of Claude after the call ends. The tool currently works only with Google Meet. It is available as a beta for customers on the Claude Team and Enterprise plans, and it is enabled by default for Team users.
The company also updated its interactive connectors. Users can now open full versions of supported third-party applications, such as Adobe, Box, and Notion, directly alongside a conversation. Claude can then make changes to the embedded application when instructed, like editing a document or retouching an image.
Industry-specific automation
For the industrial sector, Anthropic introduced Claude for Manufacturing. Skill Sprinters (opens in a new tab) details that this plugin contains 17 skills and four connectors designed to automate tasks like quote drafting, quality reporting, and supplier communication.
The connectors link Claude to Autodesk Fusion, Infor, Microsoft 365, and Oracle NetSuite. The plugin reads data from these systems to draft documents for human review. For systems without a direct connector, companies can set up regular data exports to Microsoft 365 for Claude to read. The automated skills end with a human check, meaning a responsible person must approve the work before it goes to a customer or supplier. Claude can link its numbers back to the source documents to make this verification easier.
The release lacks clear pricing details and availability information for the European Union. The sources also show conflicting information regarding which subscription tier is required to install the plugin, with mentions of both Team and Enterprise requirements.
What it means for your business
For a mid-sized company, the self-hosted agents offer a clear path to automate internal systems without handing the keys to an external vendor. A technology lead can now build agents that fix code or triage alerts while keeping the required access credentials strictly inside the company perimeter. Finance leads should note that each self-hosted deployment serves one tenant at one trust level. Running separate workloads requires separate databases and workers, which increases infrastructure costs and maintenance duties.
The meeting integration provides immediate utility for remote and hybrid teams. Having an assistant that pulls application data into a live video call saves time spent switching between tabs and searching for context. Because the feature is on by default for Team plans, operators need to establish clear rules about when the assistant should be invited to a call, especially when external guests are present.
The manufacturing plugin serves as a blueprint for specialized automation. Even if a business does not use the specific supported software, the concept of linking an assistant to operational data to draft quotes and reports is highly adaptable. It shows that the value of these tools lies in connecting them to the systems where the business already stores its data.
What to watch
Watch the development of the self-hosted agent ecosystem. Anthropic states that nothing migrates from its hosted service to the self-hosted version. Customers must rebuild their agents and memory stores from scratch. The operational burden of securing the server, managing the database, and installing updates now falls entirely on the customer. Some hosted features, including scheduled deployments and webhooks, are unavailable in the self-hosted version.
Monitor the expansion of Claude Meetings to other video conferencing platforms. The current limitation to Google Meet leaves out organizations relying on Microsoft Teams or Zoom.
Pay attention to regulatory clarity regarding the manufacturing plugin in Europe. Until Anthropic confirms data processing locations and compliance with regional rules, European operators should test the tool carefully before connecting it to live production data.
By offering options for strict data control alongside active meeting participation, Anthropic provides practical tools that businesses can mold to their own security requirements and daily workflows.