agent-deck

Agent Deck is a terminal-based command center that brings order to managing multiple AI coding agent sessions. Whether you're running Claude Code on several projects, OpenCode on others, or juggling various AI agents simultaneously, Agent Deck provides complete visibility and control from a single interface.

Key capabilities include:

• Session Management — View all running, waiting, and idle agents at a glance with instant status updates. Switch between any session with a single keystroke, eliminating the need to hunt through multiple terminal windows.
• Fork Sessions — Explore different approaches without losing context. Fork any conversation instantly while preserving the full conversation history, then fork your forks to branch out as needed.
• MCP Manager — Attach and toggle MCP servers per project or globally without editing configuration files. Changes apply automatically with session restarts handled seamlessly.
• Skills Manager — Manage Claude skills per project through a pool-based workflow. Apply changes to write project state and materialize skills into the correct directories.
• MCP Socket Pooling — Share MCP processes across sessions via Unix sockets to reduce memory usage by 85-90%, with automatic recovery from crashes in seconds.
• Advanced Search — Use fuzzy search with status filters to find sessions by running, waiting, idle, or error states. Global search across all Claude conversations is also available.
• Per-Group Configuration — Override Claude config directories and environment files per group or conductor, enabling multi-account setups within a single profile.
• Keyboard Navigation — Two-tier keybinding system for both global and group-scoped movement through the session list.

Agent Deck runs on macOS, Linux, and WSL, built with Go for performance and reliability. The tool integrates seamlessly with existing Claude and OpenCode workflows while providing the organizational structure needed for complex multi-agent projects.