When should I use Agent Teams instead of a single agent?

Use teams for complex tasks requiring parallel exploration—like simultaneous UI and API work—or when a project involves more than three genuinely parallel workstreams.

What is the difference between Agent Teams and Subagents?

Subagents report only to the caller and provide summarized results, whereas Agent Teams operate independently, communicate directly with one another via a mailbox, and manage a shared task list.

Do Agent Teams share conversation history?

No, each teammate operates in a separate context window. They load local project documentation like CLAUDE.md but do not inherit the Lead's specific conversation history unless explicitly shared via messages.

How do I enable Agent Teams in Claude Code?

Agent Teams are currently experimental and must be enabled by setting the 'CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS' environment variable to '1' in your settings.json file.

Agent Teams for Claude Code

Name: Agent Teams for Claude Code
Author: nthplusio

bynthplusio

0•

Collaboration Tools

Orchestrates multiple independent Claude Code sessions to perform parallelized research, development, and debugging through direct agent-to-agent coordination.

Agent Teams enables a sophisticated multi-agent architecture where a Lead agent coordinates specialized teammates—each operating in their own independent context window—to tackle complex engineering tasks. Unlike standard subagents, team members can message each other directly, manage a shared task list, and work on parallel workstreams such as simultaneous frontend and backend development. This skill is essential for large-scale projects that benefit from diverse technical perspectives, competing debugging hypotheses, or distributed task execution without the constraints of a single context window.

Key Features

01Multi-agent orchestration with dedicated Team Lead and specialized Teammate roles

02Independent context windows for teammates to prevent context bloat and enable parallel work

03Shared task management system for tracking, claiming, and updating project progress

04Direct agent-to-agent mailbox system for peer-to-peer communication and coordination

05Pre-designed blueprints for research, feature development, code review, and planning phases

060 GitHub stars

Use Cases

01Parallel feature development where agents independently own frontend, backend, and testing modules

02Complex debugging and security audits where teammates test competing hypotheses in parallel

03Strategic planning and research sessions that require synthesis of multiple technical perspectives

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nthplusio/functional-claude agent-teams

For use in Claude.ai and ChatGPT

Download Skill