# SEE CMUX SOLVE Multi-Agent Orchestration (Claude Code and Pi Agent)

## Executive summary

This video explores advanced multi-agent orchestration patterns and the critical need for 'agentic access' to improve agent performance. The speaker argues that simply running many agents in a loop ('vibe coding') is insufficient; true progress requires visibility and programmatic control over the entire process. Cmux is presented as a tool that solves three major problems: lack of programmatic access, inability to monitor/improve agents, and slow manual setup time for large agent teams. The core architectural pattern recommended is three-tier orchestration (Orchestrators prompt Leads, who prompt specialized Experts).

## Key takeaways

- Agentic Access is Paramount: An agent you cannot see or programmatically interact with cannot be improved. This principle distinguishes 'agentic engineering' from simple 'vibe coding.'
- Three-Tier Orchestration Pattern: The recommended structure for multi-agent teams is: Orchestrators prompt the Leads, and Leads prompt specialized Agent Experts.
- Scaling Compute to Scale Impact: Multi-agent orchestration allows scaling compute by running multiple diverse agents (e.g., Claude Code, Codex, Pi agent) in parallel on the same problem, ensuring visibility into all attempts.

## Technical details

- Cmux Functionality & Architecture: Cmux provides programmatic control over terminal surfaces (panes and windows), allowing agents to send information (send key), read screen output after actions (read the screen), and manage surfaces (open/close).
- Multi-Agent Orchestration Problems Solved: 1. No programmatic access: Cmux provides agentic access to every terminal. 2. Monitoring/Improvement: Allows observation of all agent actions for refinement via core four (re-emphasize behavior or present negative rewards). 3. Slow Setup: Features reusable session files and agentic access enable quick launch of large teams.
- Agent Team Scaling Example: Demonstrated booting an eight-agent 'red team' in parallel to solve a problem (e.g., security vulnerability search), showcasing how different models (Opus 4.8, Sonnet, GLM 5.2) contribute simultaneously.
- Cmux vs. Tmux: While tmux is a tried-and-true tool, Cmux offers superior programmability and customizability for the modern agentic workflow, making it ideal for building complex developer experiences (DX).

## Practical implications

- Implement a three-tier agent architecture (Orchestrator -> Leads -> Workers) for complex tasks to maintain structure while allowing flat communication.
- Prioritize tools with robust programmatic access and API capabilities over those that only offer visual interfaces, as this enables true 'agentic speed.'
- Use multi-agent fleets for critical tasks like hotfix deployment or security vulnerability scanning, ensuring all specialized models contribute in parallel to maximize the chance of a rapid solution.

## Topics

Agentic Engineering, Multi-Agent Systems, DevOps Tooling, Programmatic Control, Terminal Multiplexing, Cmux Codebase, Cmux Official Site, Five Pillars of Agentic Engineering

Source: https://www.youtube.com/watch?v=WAFUMBLOjHo