What changed
this month.
This is not a tech changelog. It's the small, short list of real Claude and Anthropic updates that are actually worth five minutes of a busy owner's attention, and nothing else. Updated as things happen, not on a schedule.
Claude Cowork is rolling out as the easiest way to delegate real work to Claude.
Cowork is a desktop app where you point Claude at a folder on your computer, describe an outcome, and it plans the work and asks you to approve before it executes. For a small business, this is the first version of Claude built for someone who isn't a developer and doesn't want to learn a terminal. It runs in an isolated space on your machine, so it can only touch what you give it access to.
Connectors for everyday business tools (Gmail, Google Drive, Notion, Slack) got easier to turn on and scope down.
Connectors let Claude read and act inside the tools you already use instead of you pasting things in by hand. The newer versions let you limit a connector to one folder, one label, or one channel instead of granting access to everything. If you've been curious about this but weren't sure where to start, this is the capability worth trying first. See our Blueprint article on MCP and connectors for the plain-English walkthrough.
Claude in Slack moved from a limited beta to general availability for Team and Enterprise workspaces.
You can now @-mention Claude directly inside a Slack channel or DM and get it to summarize a thread, draft a reply, or answer a question using context from that conversation. For small teams that live in Slack day to day, this cuts out a step, you no longer have to copy a conversation into a separate chat window to get help with it.
Anthropic expanded the Claude model lineup with clearer guidance on which size to use for which job.
Anthropic's current lineup spans a small, fast model for simple tasks, a mid-size model for everyday writing and analysis, and a large model for hard reasoning and long documents. The practical takeaway for a business owner: you're very likely overpaying for compute if you're using the largest, most expensive model for tasks like sorting emails or quick summaries that the smaller model handles just as well.
Don't see something you expected here? We leave out routine changelog items and anything we're not confident is accurate. If you think we missed something that matters, let us know.