asheshgoplani/agent-deck

What it solves

Agent Deck provides a centralized command center for managing multiple AI coding agents (such as Claude Code and OpenCode) simultaneously. It eliminates the need to juggle numerous terminal windows by offering a single interface to track, switch between, and orchestrate a fleet of agent sessions across different projects.

How it works

It operates as a Terminal User Interface (TUI) and CLI that manages agent sessions, often utilizing tmux for session isolation. It integrates with various AI tools and provides a management layer for configuration, environment variables, and project state. It also includes a "conductor" system that can be linked to communication platforms like Telegram to supervise and notify the user about agent progress.

Who it’s for

Developers who use multiple AI coding agents across various repositories and need a way to organize, monitor, and switch between these sessions efficiently without losing context.

Highlights

• Session Management: Quick switching, forking of conversations, and status detection (Running, Waiting, Idle, Error).
• Conductor System: A supervising agent that can be controlled via Telegram, Slack, or Discord to monitor other worker agents.
• MCP & Skills Manager: Easy attachment of Model Context Protocol (MCP) servers and Claude skills without manual config file editing.
• Git Worktree Integration: Support for isolated working directories per agent to prevent conflicts, including automated setup/destruction scripts and support for bare repositories.
• Resource Optimization: An MCP Socket Pool that shares MCP processes across sessions to reduce memory usage by 85-90%.
• Account Switching: Ability to migrate existing conversations between different AI accounts on the fly.