Claude Code's Agent Teams and Open Orchestrator solve different problems but can work together powerfully. This guide explains the differences and integration patterns.

Quick Comparison

Feature	Agent Teams	Open Orchestrator
Scope	Same codebase, multiple agents	Multiple branches, multiple codebases
Coordination	Agents message each other	Control Plane + CLI commands
Isolation	Shared git worktree	Separate worktrees with independent environments
Dependencies	Same node_modules/venv	Each worktree has its own dependencies
Best For	Code review, competing hypotheses, research	Parallel feature development, branch management
Infrastructure	Built into Claude Code	CLI + git worktrees + tmux

The Key Difference

Agent Teams: Intra-Branch Collaboration

Multiple AI agents working together in the same codebase:

┌─────────────────────────────┐
│   feat/auth branch          │
│                             │
│   Agent 1: Security Review  │
│   Agent 2: Testing          │
│   Agent 3: Documentation    │
│                             │
│   All working on same files │
└─────────────────────────────┘

Use when:

• You want agents to debate different approaches
• You need multiple perspectives on the same code
• You want agents to review each other's work
• You're exploring competing hypotheses

Open Orchestrator: Cross-Branch Orchestration via the Control Plane

The Control Plane is Open Orchestrator's answer to agent teams -- a multi-provider cockpit that supervises Claude Code, Pi, Droid, and OpenCode across git worktrees. Instead of multiple agents in one branch, you get multiple isolated environments working in parallel, all visible from a single prioritized control plane:

open-orchestrator · 3 rows · 14:32:08

IN FLIGHT
  feat/auth   Claude Code   Own deps · Own .env
  feat/api    OpenCode      Own deps · Own .env
  fix/login   Droid         Own deps · Own .env
        ↑ Control Plane (owt)

Use when:

• You are working on multiple features simultaneously
• You need complete isolation between tasks
• You want different branches with different dependencies
• You want to use different AI tools for different tasks

Integration Patterns

The real power comes from using both together. Here are proven patterns:

Pattern 1: Agent Teams per Feature Branch

Create isolated worktrees with Open Orchestrator, then spawn Agent Teams inside each:

# Create isolated worktree for feature A
owt new "Implement OAuth authentication"

# Inside that worktree, spawn Agent Team
# - Agent 1: Implement OAuth flow
# - Agent 2: Write tests
# - Agent 3: Security audit

# Meanwhile, create another worktree for feature B
owt new "Build payment gateway"

# Monitor both from the Control Plane
owt

Why this works:

• Complete isolation between features (no merge conflicts)
• Each team has its own dependencies/environment
• Teams work in parallel without interference
• Monitor all teams from the Control Plane

Pattern 2: Parallel Experiments

Explore multiple approaches simultaneously:

# Create worktrees for different experiments
owt new "Try REST API implementation" -t feature
owt new "Try GraphQL implementation" -t feature
owt new "Try gRPC implementation" -t feature

# Each worktree works in isolation with its own dependencies
# Compare results across worktrees without conflicts

# See all experiments in the Control Plane
owt

Why this works:

• Try multiple architectures without contamination
• Each approach has isolated dependencies
• Easy to compare and merge winning approach
• No cleanup needed for rejected experiments

Pattern 3: Infrastructure + Collaboration

Use Open Orchestrator for infrastructure, Agent Teams for collaboration:

# Use Open Orchestrator to create isolated environment
owt new "Refactor auth module"

# Inside, use Agent Teams for specialized tasks:
# - Agent 1: Security review
# - Agent 2: Performance optimization
# - Agent 3: Test coverage

# Send tasks from the Control Plane
owt send feat/refactor-auth-module "Review the authentication module"

When to Use What?

Use Agent Teams When:

• Multiple perspectives needed on same code
• Agents should debate and critique each other
• Exploring competing solutions to one problem
• Code review, security audit, research tasks
• Working in a single branch

Use Open Orchestrator When:

• Working on multiple features simultaneously
• Need complete isolation between tasks
• Different branches need different dependencies
• Coordinating work across different tasks
• Want to use different AI tools for different tasks

Use Both Together When:

• Multiple teams working on different features
• Each feature needs agent collaboration
• You want both isolation and coordination
• Managing complex multi-branch projects
• Running parallel experiments with agent teams

Complementary Strengths

What	Agent Teams	Open Orchestrator
Agent coordination	Built-in messaging	Control Plane + owt send
Task list sharing	Shared todo list	Status tracking per worktree
Branch isolation	Same branch	Different branches
Environment isolation	Shared deps	Independent deps
Infrastructure management	Not provided	tmux, git worktrees, status
Visual overview	Not provided	Prioritized control plane

Summary

Use Agent Teams when you need agents to collaborate on the same code.

Use Open Orchestrator when you need isolated environments for different tasks.

Use both together for maximum parallelism: agent teams working within individual branches, all managed from the Control Plane.

Next Steps

• Learn about the Control Plane
• Explore Parallel Features Workflow
• Learn about Status Tracking