Anthropic adds multi-agent Projects to Claude Code (beta)

Anthropic announced a redesigned Projects experience for Claude Code on September 17, 2026, releasing the update in beta with a coordinator that directs parallel agent threads drawing on shared memory and a common library of files and artifacts. The company said managing multiple sessions across a build used to require users to split the work themselves, manage handoffs manually, and recombine the results at the end, but in the updated experience, the user describes what needs to get done and Claude manages the work.

The new beta experience for Claude Code projects lets one project coordinate parallel threads, shared memory, and a project library to manage long-running work across repos and tasks with less handoff and more progress steering. This represents a fundamental shift in how developers interact with AI-powered coding assistants, moving from managing individual agent sessions to orchestrating multiple concurrent workflows through a central hub.

From Manual Management to Automated Coordination

The redesigned Projects feature transforms Claude Code from a single-conversation coding tool into a multi-threaded orchestration platform. Previously, developers working on complex builds needed to manually divide tasks among different Claude sessions, track progress across multiple conversations, and then integrate the results themselves. The cognitive overhead of this approach limited the complexity and scale of projects that could be effectively managed with AI assistance.

Under the new system, developers provide high-level descriptions of what needs to be accomplished, and Claude Code’s coordinator automatically divides work among parallel agent threads. These threads share access to a common project library containing files, artifacts, and context, eliminating the need for repetitive information handoffs between sessions.

Shared Memory and Project Libraries

The shared memory architecture allows all agent threads within a project to draw on the same contextual foundation. When one thread learns something new about the codebase, discovers a bug pattern, or generates a useful artifact, that information becomes immediately available to other threads working on related tasks. This shared knowledge base reduces redundant work and helps maintain consistency across parallel development streams.

The project library serves as a centralized repository for files and artifacts that multiple agents need to reference. Rather than uploading the same documentation or code samples to each individual conversation, developers can populate the project library once and allow all agent threads to access these resources as needed.

Availability and Access

The redesigned Projects experience is currently available in beta for users on Claude Pro and Max plans. According to coverage of the announcement, the feature also supports GitHub repositories through integration with the Claude GitHub app, allowing developers to connect their existing codebases directly to the multi-threaded coordination system.

The timing of this release positions Anthropic to compete more directly with other AI coding platforms that offer multi-agent workflows. By automating the orchestration that developers previously handled manually, Claude Code aims to make complex, long-running development projects more manageable through AI assistance.

Technical Implementation

While specific technical details about the coordinator’s decision-making process have not been disclosed, the system appears to function as an intelligent task router that evaluates project requirements and allocates work to appropriate agent threads. The coordinator must balance multiple considerations, including task dependencies, resource constraints, and optimal parallelization strategies.

The shared memory implementation raises questions about context window management and how the system prevents information overload as projects grow. Traditional single-session AI coding tools struggle with maintaining context over long development cycles, a challenge that becomes more complex when multiple parallel threads all draw from the same memory pool.

Implications for Development Workflows

The shift toward agent coordination rather than single-session coding assistance reflects broader industry trends in AI-powered development. As language models become more capable of sustained, complex work, the bottleneck increasingly lies in orchestrating multiple AI agents rather than in the capabilities of individual models.

For development teams, this could mean rethinking how they structure AI-assisted workflows. Instead of assigning discrete coding tasks to individual AI sessions and integrating the results manually, teams might describe entire feature requirements or refactoring initiatives and allow the coordinator to manage the subdivision and parallel execution.

The redesigned Projects feature also has implications for how developers learn to work with AI coding assistants. Rather than focusing on prompt engineering for individual tasks, effective use of the new system may require developing skills in project description, requirement specification, and high-level coordination rather than low-level task management.

Key Facts

Anthropic announced the redesigned Projects experience for Claude Code on September 17, 2026
The beta feature coordinates parallel threads, shared memory, and a project library for managing long-running work across repos and tasks
The feature is available on Claude Pro and Max plans
Integration with GitHub repositories is supported through the Claude GitHub app
Previously, managing multiple sessions required users to divide work, manage handoffs manually, and recombine results themselves