Anthropic has introduced a new projects feature for Claude Code, allowing users to run multiple parallel AI agent sessions within a single workspace.
Key facts
- •The projects feature allows multiple Claude Code sessions to run in parallel on separate git branches.
- •A coordinator manages thread delegation, output review, and the assembly of final results.
- •The system can handle complex tasks like migrations or work spanning multiple repositories.
- •Anthropic warns that running multiple concurrent threads can lead to rapid token consumption.
- •The feature is currently in beta for select Claude Pro and Max subscribers.
Anthropic has launched a beta version of its new projects feature for Claude Code, designed to enable users to manage multiple related tasks simultaneously. The system allows users to run several AI agent sessions in parallel, with a coordinator managing the workflow, delegating tasks, and assembling final results. This update is currently available to select Claude Pro and Max subscribers, with plans to expand access to Team and Enterprise users in the future.
How Parallel Sessions Work
The new projects feature functions by spinning up individual Claude Code cloud sessions for each task, with each session operating on its own git branch and repository copy. A coordinator oversees these threads to keep work organized; if multiple threads modify the same code, the system handles the overlap as a merge conflict, similar to standard pull request workflows. Within these sessions, work can be further divided into smaller pieces using subagents, loops, and workflows to accelerate the completion of large assignments.
Management and Capabilities
Users can interact with individual threads or monitor progress through a main project chat interface. Anthropic notes that the system can manage tasks even after a user steps away from their computer, and progress can be steered from mobile devices. The company has introduced five distinct ways to run multiple tasks, including Agent View for overseeing independent tasks, Agent Teams for coordinating instances that share information, and Dynamic Workflows for script-orchestrated subagents. While currently limited to cloud sessions, Anthropic stated that support for local tools and code is expected to arrive soon.
Resource Usage and Availability
Anthropic has cautioned that the projects feature can consume tokens rapidly because each thread operates as a full Claude Code session. The feature is distinct from previous iterations of projects that focused on self-contained workspaces with separate chat histories. Following the initial beta release for select Pro and Max subscribers, the company intends to roll out access to all Pro, Max, Team, and Enterprise users, as well as Cowork and regular Claude chat users.
Advertisement
This article was independently rewritten by ManyPress editorial AI from reporting originally published by The Verge AI, The Register.
