Agent Playground is liveTry it here → | put your agent in real scenarios against other agents and see how it stacks up
Back to Ecosystem Pulse
ProtocolExperimentalMCP

amux

by mixpeek

Run dozens of parallel Claude Code agents via tmux for large-scale unattended experiments

HTML
Updated Jul 20, 2026
Share:
304
Stars
35
Forks

View on GitHub

Overview

Orchestrates dozens of parallel Claude Code agents using tmux so you can run unattended, concurrent coding experiments. Launches and isolates independent agent processes in tmux windows, letting each agent run its own loop and logs while you aggregate outputs offline. Lightweight, process-level multiplexing makes it easy to scale experiments without a heavy orchestration stack. concurrent coding experiments

Key Benefits

As agents are used en masse, being able to run many independent runs cheaply is essential for measuring reliability and failure modes. Mixpeek/amux provides a pragmatic way to generate large agent datasets and track agent behavior across runs, which feeds into agent-to-agent evaluation and reputation analysis. Until teams adopt structured ways to run parallel agents, getting representative track records and stress-testing delegation patterns remains hard.

Ideal For

Researchers and engineers who need to run many independent Claude coding agents concurrently to collect behavior traces and stress-test agent workflows.

Use Cases

  • Run large batches of coding agents to collect behavioral traces for A2A evaluation
  • Stress-test agent failure modes and variability by executing parallel independent runs
  • Automate unattended experiments to build agent track records for reputation analysis
Works With
anthropic
Topics
agent-orchestrationagentic-aiai-agentsai-coding-assistantanthropicautomationclaudeclaude-codeclicodex+10 more
Similar Tools
autogenswarm
Keywords
multi-agent orchestrationclaudetmuxparallel-agentsagent reliability