Anthropic Upgrades Claude Code Projects to Run Parallel AI Workers Hands-Free
Claude Code Projects coordinator is now proactive, answering questions and pushing thread status updates in chat. Individual threads can be shared as read-only links to signed-in Claude users; threads with local computer access are excluded.
Takeaways
- Claude Code Projects coordinator is now proactive, answering questions and pushing thread status updates in chat.
- Individual threads can be shared as read-only links to signed-in Claude users; threads with local computer access are excluded.
- Reminders let a thread hide itself and resurface in the project chat at a scheduled time.
- /btw opens a side thread for tangents; /fork branches a new thread from the current one.
- Threads panel adds a recents row and artifact thumbnails for easier retrieval across large projects.
- Public beta on Pro and Max across web, desktop, and the latest mobile app; Team and Enterprise excluded.
Anthropic has updated Claude Code Projects with proactive coordinator messages, read-only thread sharing, reminders, branching commands, and improved artifact navigation. The changes reduce the monitoring required when several cloud sessions work in parallel.
Each project is a long-running conversation with a coordinator that assigns work to threads. Every thread runs as an independent Claude Code cloud session with its own repository copy and Git branch. The coordinator directs the project from thread reports, while overlapping code changes surface as standard Git merge conflicts when branches are combined.
The coordinator reports without prompting
The coordinator now answers questions directly in project chat and posts status updates as threads progress. Previously, developers had to monitor the thread list to find sessions waiting for input. The updated behavior brings pending decisions, completed tasks, and blocked work into the main conversation.
Read-only links stop at local access
Individual threads can now be shared through read-only links with anyone signed in to Claude. A developer can open the thread menu, select Share , and send the generated URL without exposing the rest of the project.
Threads that accessed a local computer remain unshareable. The restriction prevents shared transcripts from exposing local file contents, shell output, or other machine data gathered during the session.
Park a thread or branch the work
| Control | Behavior | Best use |
|---|---|---|
| Reminders | Hide a thread until a specified time, then return it to the thread list and notify the coordinator. | Work that depends on a later review, deployment, or external response. |
| /btw | Open a separate thread for a tangent while preserving the current thread’s goal and context. | Investigating a related issue without diverting active work. |
| /fork | Create a new thread from the current thread’s state. | Testing an alternative implementation or debugging path. |
The two slash commands were already available in regular Claude Code sessions. Their addition to Projects brings the same branching workflow to coordinated, multi-thread work.
Artifacts surface sooner
The threads panel now places recently opened, updated, and pinned threads in a row at the top. Each thread can also display thumbnails for generated documents, diagrams, and other artifacts, with the previews controlled through view settings.
The redesign simplifies project settings and local-folder access as well. Anthropic has released the changes on the web, desktop clients, and the latest mobile app.
Parallelism draws from the same allowance
Anthropic introduced the public beta on September 17 for Pro and Max subscribers who joined the waitlist. Team and Enterprise plans remain unsupported. Eligible subscribers who lack access can still join the waitlist .
Projects carry no separate fee, but every worker consumes the subscriber’s existing Claude Code allowance. Anthropic limits accounts to 200 new threads per day across all projects. A thread that reaches a usage limit pauses and resumes automatically when capacity becomes available.
Project usage appears separately in the usage controls, and developers can choose the model and effort level for the coordinator and worker threads. New projects default to Opus for both roles, with low effort for the coordinator and high effort for workers. Several high-effort sessions running at once can exhaust plan limits faster than a single Claude Code session.
Where many threads pay off
Projects fit engineering work that can be divided into independent tasks and reconciled through branches and pull requests. Anthropic’s examples include:
- Performance work: Assign separate threads to profile endpoints, test optimizations, and open pull requests. A target such as p75 checkout latency refers to the response time at or below which 75% of requests finish.
- Cross-repository migrations: Connect API, web, and mobile repositories, then create one thread per codebase and report the required merge order.
- Batch changes: Distribute refactors, dependency upgrades, test repairs, or bug-triage items across parallel sessions.
Tasks with strict dependencies still require developers to sequence pull requests and resolve conflicts. Projects also remain cloud-only and limited to Claude Code, while parallel workers draw from the same plan limits. Small changes that fit one branch and one context window generally consume fewer resources in a regular Claude Code session.
Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.
Read original report
