Claude Can Now Juggle Several Coding Jobs at Once
Anthropic has changed Projects in Claude Code so several AI helpers can work at the same time using the same memory, the same goals, and the same set of files. The new version is in beta for some Claude Pro and Max subscribers, The Verge reported on Sept. 17.
Each project can run several threads, meaning separate jobs, in parallel. A coordinator hands out the work and tracks it. You can open any single thread to work with it directly, or stay at the project level and check progress and give instructions in the main project chat.
Each thread is a Claude Code session that runs in the cloud. It gets its own branch, which is simply its own safe copy of the code. That separation lets each helper work without getting in the way of the others. Differences are sorted out later when the work is combined. It is a bit like a kitchen where each cook gets their own counter, and a head chef brings the dishes together at the end.
If two threads change the same piece of code, that overlap is fixed as a merge conflict, handled like a pull request, which is the normal check engineers use before combining code. A thread can also split its own job into smaller pieces by handing work to smaller helpers called subagents, and by running repeated steps and set routines under the coordinator.
At launch, threads run only in the cloud. Support for local tools and local code on your own computer is coming very soon. Anthropic plans to open Projects to all Pro, Max, Team, and Enterprise users, as well as to Cowork and regular Claude chats.
The broader change here is for the person doing the work. Older single helpers left you to break jobs apart, put them in order, and fit the results together. This system does more of that organizing, so you spend more time checking results and giving direction.
In my view, using separate copies of the code and the normal combine-and-check steps is a down-to-earth choice for engineers. They already use those steps every day, so there is nothing new to learn, and the AI output fits into the testing and checking they already do.
What to watch now is how it behaves day to day. How often helpers step on each other's work, how well the shared memory holds up, and how well the coordinator keeps track as you add more threads will decide whether it feels like a coordinated team or a pile of competing drafts. Being able to jump in on one thread or from the main chat also makes you both manager and checker.
Looking ahead, the benefit if that extra managing stays easy is simple. Small teams could keep several jobs moving without jumping back and forth, while the AI tries out versions and people focus on what they want, how parts connect, and what counts as finished. That would make coding work calmer and more shared.
