Glossary

Key concepts in AI agent orchestration and parallel development.

• Agent Multiplexer
  Agent Multiplexer: A tool that runs and manages multiple AI coding agents simultaneously. → Best AI agent multiplexers 2026
• Self-Healing AI Agent
  Self-Healing AI Agent: An AI agent that automatically recovers from crashes, context exhaustion, and stuck states. → Self-healing configuration guide
• Context Compaction
  Context Compaction: The process of summarizing conversation history to free up context window space in an AI agent. → How amux handles context compaction
• YOLO Mode
  YOLO Mode: A configuration that auto-approves tool-use prompts in Claude Code for fully unattended operation. → Setting up YOLO mode
• Agent Orchestration
  Agent Orchestration: Coordinating multiple AI agents to work together on complex tasks. → AI agent orchestration guide
• Parallel Coding Agents
  Parallel Coding Agents: Multiple AI agents working on different coding tasks at the same time. → Running 10+ agents guide · Agentmaxxing
• Token Accounting
  Token Accounting: Tracking token usage and costs across multiple AI coding sessions. → Token spend tracking guide · Cost calculator
• Atomic Task Claiming
  Atomic Task Claiming: A concurrent-safe mechanism for agents to claim tasks without duplicating work. → Kanban board guide
• Session Persistence
  Session Persistence: The ability for AI agent sessions to survive restarts and maintain their identity. → Session memory guide
• Prompt Replay
  Prompt Replay: Automatically re-sending the last message after an AI agent crash to recover progress. → Self-healing configuration
• Global Memory
  Global Memory: A shared knowledge base accessible to all AI agents in a multiplexer. → Agent-to-agent orchestration
• Background Watchdog
  Background Watchdog: A monitoring process that detects and recovers from AI agent failures automatically. → Configure the watchdog · Running agents overnight
• Conversation Fork
  Conversation Fork: Cloning an AI session's history into multiple new sessions for parallel exploration. → Conversation forking guide
• Terminal Peek
  Terminal Peek: Viewing an AI agent's live terminal output without attaching to the session. → Debugging agent fleets
• PWA (Progressive Web App)
  PWA (Progressive Web App): A web application installable on mobile devices with offline support. → Mobile management guide
• tmux Session Management
  tmux Session Management: Using tmux terminal multiplexer to run persistent background processes. → amux vs DIY tmux scripts
• REST API Orchestration
  REST API Orchestration: Using HTTP endpoints to coordinate AI agent communication and task management. → REST API reference
• iCal Sync
  iCal Sync: Exporting task deadlines as calendar events via the iCal format. → Kanban board guide
• SQLite-Backed Board
  SQLite-Backed Board: A kanban board using SQLite for concurrent-safe task management. → Kanban board guide
• Git Worktree Isolation
  Git Worktree Isolation: Using git worktrees to give each AI agent its own working copy of a repository. → Git conflict avoidance guide
• Headless Coding Agent
  A coding agent that runs without a GUI, controlled entirely via CLI or API. → Claude Code headless mode guide
• Vibe Coding
  A programming style where developers describe intent in natural language and AI generates the implementation. → Vibe coding quality gates · Vibe debugging guide
• Agent Fleet
  A group of AI coding agents running in parallel, coordinated toward a shared goal. → Agentmaxxing guide · Scaling to 50+ agents
• LLM Context Window
  The maximum amount of text an LLM can process in a single request — a key constraint in long-running agent sessions. → Context engineering for agents
• Unattended Coding
  Running AI coding agents without human supervision — enabled by self-healing, YOLO mode, and atomic task management. → Running agents overnight · Setting up YOLO mode