What Claude Code's New Projects Means for Your Business

Anthropic has relaunched Claude Code Projects, giving developers an always-on conversation that remembers context and manages multiple AI coding agents running in the cloud at once, rather than starting each chat from scratch.

What it does

Instead of opening a new chat every time you need a fix, a feature or a test written, you open one Project. Claude remembers what the codebase looked like last time, tracks several jobs running in parallel in the cloud, and reports back when each one is done or stuck. You are not babysitting a single chat window all day, you are checking in on a queue.

What it means for your hours

If your firm runs any in-house software, a stock system for a parts business, a job-tracking tool on site, a scheduling app, this changes how a developer's day works. Say you need three things done this week, fix a sync bug in the stock system, build a supplier import script and patch a bug on the customer portal. Under the old model, a developer explains context three separate times, in three separate chats, and checks each one manually. Under Projects, all three can run inside one thread that already knows the codebase, working in parallel while the developer does something else.

The saving is not the AI writing the code, that part already existed. The saving is the time your developer used to spend re-explaining context and switching between disconnected sessions. For a small software team, that is real hours back each week, hours that can go on client work instead of housekeeping.

The risk sits in the word delegates. If nobody reviews what several agents did overnight, mistakes compound quietly. Anthropic's own tooling flags when an agent is stuck, but a human still needs to check the change before it goes live.

What to do this month

• If you already pay for Claude Code, trial Projects on one real piece of ongoing work rather than a toy task.
• Set a rule that someone reviews every change before it merges, however small.
• Track the hours your developer would normally spend re-explaining context, and compare it after a fortnight on Projects.