Anthropic launches Claude Code Projects for multi-agent workflows

Anthropic has introduced a revamped "Projects" feature in Claude Code, enabling users to manage teams of AI agents that coordinate on complex software tasks. According to The Verge, this update allows developers to run multiple agents under a single roof, sharing memory, goals, and a library of files and artifacts. The launch marks a significant shift from single-agent interactions to orchestrated multi-agent workflows, similar to tools like Grok Bot that manage groups of AI agents.

How the multi-agent system works

The core of the new Projects feature is the use of "threads," which are individual Claude Code cloud sessions working on their own branches and copies of the repository. A central "coordinator" agent directs these threads, keeping the overall work organized. If different threads modify the same code, the system resolves the overlap as a merge conflict, treating it just like any other pull request. This approach allows for parallel development without immediate human intervention for every step.

Each thread can further delegate work using subagents, loops, and workflows, allowing large assignments to be broken down and finished faster. Users can interact with each thread individually or monitor and update the entire operation through a main project chat. This architecture mirrors traditional software engineering practices where a tech lead coordinates multiple engineers working on different parts of a codebase.

Availability and access

The updated Projects feature is available in beta starting today for select Claude Pro and Max subscribers. Anthropic plans to expand access to all Pro, Max, Team, and Enterprise users later. The company also indicates that the feature will eventually be available in Cowork and regular Claude chats. Currently, threads run in the cloud, but Anthropic states that support for local tools and code is coming "very soon." This phased rollout suggests Anthropic is testing the reliability and performance of the multi-agent coordination before making it widely available.

Context in the AI devtools landscape

This launch positions Claude Code more aggressively in the competitive space of AI coding assistants. While many tools focus on single-agent code generation or completion, the ability to coordinate multiple agents on parallel tasks is a differentiator. The use of a coordinator agent to manage merge conflicts and task delegation aligns with the growing trend of AI systems that mimic human team dynamics. This is particularly relevant for large codebases where a single agent might struggle to maintain context or handle complex, multi-file changes efficiently.

What it means for developers

For developers, this feature offers a new way to tackle large-scale refactoring, feature development, or bug fixing. By allowing multiple agents to work in parallel, projects that previously required sequential steps can now be executed concurrently. However, developers need to be careful with the merge conflict resolution process, as automated merges can sometimes introduce subtle bugs. The ability to interact with individual threads provides a level of control that can help mitigate these risks. This feature is particularly useful for teams using Claude Pro or Max, as it leverages the higher limits and capabilities of these tiers.

What to watch

• Local execution support: Anthropic has promised support for local tools and code "very soon," which will be crucial for developers who prefer or require local execution for security or performance reasons.
• Expansion to all users: The timing of the expansion to all Pro, Max, Team, and Enterprise users will indicate how confident Anthropic is in the stability of the feature.
• Merge conflict handling: Real-world performance of the automated merge conflict resolution will be a key metric for adoption, as it directly impacts code quality and developer trust.