Anthropic on Thursday released in beta a redesigned Projects system for Claude Code in which a coordinator splits a goal into parallel threads, each a full Claude Code cloud session. Claude coordinates from the project conversation, while every worker runs on Opus at high effort by default and reports back when it finishes. Each thread draws on the subscriber’s plan, so projects can reach usage limits faster.

What Changed

AI-generated summary, reviewed by an editor. More on our AI guidelines.

A coordinator above the threads

Each worker gets its own branch and copy of the repository. A thread can divide its assignment further through subagents, loops or workflows. If workers edit the same code, Projects surfaces the overlap as a merge conflict like any other pull request.

Shared memory carries decisions between sessions. The notes are stored as files, with a MEMORY.md index that every new thread reads. Users can inspect, edit or delete those files in Project settings. A library holds uploaded material and artifacts produced during the work.

The system keeps running after a laptop closes and can be steered from a phone. The launch examples include profiling endpoints to reduce checkout p75 latency and retiring a deprecated v1 endpoint across separate API, web and mobile repositories. In the second case, Claude opens a thread for each repository and reports which pull requests must merge first.

Opus is the default

The coordinating conversation uses low effort. Anthropic’s documentation says the thread default “draws on your plan fastest.”

There is no fixed number of concurrent threads. A limit the user asks Claude for is saved as a preference Claude follows, not a hard cap; the only enforced ceiling is 200 new threads per day across a subscriber’s projects. An idle thread watching a pull request can wake and consume more of the plan when a test fails or a reviewer comments. A Usage tab breaks consumption down by thread and model.

Anthropic has published no fixed usage figure for an individual thread. Without paid usage credits, a thread that reaches the plan limit waits and resumes when it resets. A thread started by a routine is the exception: its turn stops with a limit error, and the user must message it after the reset.

A rollout amid a billing dispute

The beta initially covers select Pro and Max subscribers who have used cloud sessions and have no existing projects in chat or Cowork. Other users on those plans can join a waitlist. Team and Enterprise access will come later, while existing projects remain unchanged during the rollout.

Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.

The release follows an expanded class-action complaint filed on September 8, 2026 over Anthropic’s Max subscription claims. The plaintiffs allege the advertised 5x and 20x usage multipliers apply within five-hour sessions while a separate weekly ceiling restricts total use. Anthropic argued in a motion to dismiss an earlier version that buyers could reach details about the limits through links in the purchase flow.

Cloud-only work

David Gewirtz, a ZDNET senior contributing editor who uses Claude Code in a terminal with local files, said the redesign does not create new parallelism because he already uses subagent swarms. He also said overlapping work postpones conflicts instead of preventing them. Gewirtz said the launch post left memory under-documented. The fuller documentation explains where users can inspect its files.

The worker sessions currently have no access to tools or files that exist only on a user’s computer, or to services behind a private network. In projects with several repositories, permission rules, hooks and environment values from each repository’s .claude/settings.json do not apply. Anthropic says local execution is “coming very soon,” but has given no date.

Frequently Asked Questions

What are the redesigned Claude Code Projects?

A project is one conversation in which Claude acts as coordinator. It splits a goal into parallel threads, each a full Claude Code cloud session with its own branch and copy of the repository. Threads report back to the project conversation, and overlapping edits surface as ordinary merge conflicts.

Which model do project threads use by default?

A new project runs Opus everywhere. Worker threads start at high effort and the coordinating conversation at low effort. Anthropic's documentation says the thread default draws on a subscriber's plan fastest. Model and effort can be changed in project settings.

Is there a limit on how many threads a project runs?

There is no fixed number of concurrent threads. A limit the user asks Claude for is saved as a preference, not a hard cap. The only enforced ceiling is 200 new threads per day across a subscriber's projects.

What happens when a thread hits the plan's usage limit?

Without usage credits turned on, a thread waits and resumes when the limit resets. A thread started by a routine is the exception: its turn stops with a limit error and the user has to message it after the reset.

Who can use the new Projects now?

The beta covers select Pro and Max subscribers who have used cloud sessions and have no existing projects in chat or Cowork. Others on those plans can join a waitlist. Team and Enterprise access comes later, and threads currently run only in the cloud.

AI-generated summary, reviewed by an editor. More on our AI guidelines.

Anthropic Cuts Claude Code Weekly Limits 17% on Sept. 14
Claude Code users describe checking usage every 30 minutes and rationing their week as Anthropic prepares to cut current weekly limits 17% on September 14.
Implicator.ai
Astra Pulls Claude Code's Heaviest Users as Limits Drop 17%
Anthropic's 50% Claude Code boost ends September 13, and the permanent 25% raise replacing it leaves subscribers about 17% below this week's meter. GPT-6 Astra lands in the same week, fresh off an independent run through all of Valve's Portal with no human input.
Implicator.ai
Anthropic Ships /skill-doctor to Audit Claude Code Skills
Anthropic spent the summer telling developers their Claude configuration was too long. On September 4 it shipped a command that finds the excess. /skill-doctor reports which loaded skills a session never used, and what each one costs in context on every turn.
Implicator.ai
Tools & Workflows

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai