Agent Deck vs tmux for managing AI sessions?

Agent Deck builds on tmux by adding AI-specific features like smart status detection, session forking with context inheritance, and integrated MCP management. While tmux handles terminal multiplexing, Agent Deck provides organized groups, cost tracking, and tool integration tailored for AI workflows.

Can Agent Deck work with Cursor AI?

Yes, but with limitations: Agent Deck supports Cursor in terminal mode for status detection and organization, but full features like forking and MCP management are primarily designed for Claude Code. Custom configuration might be needed for deeper integration.

How to track costs for Gemini CLI in Agent Deck?

Cost tracking for Gemini CLI is implemented via output parsing, but the README notes it as untested. You can enable it in the configuration, but reliability may vary compared to Claude Code's hook-based integration, and pricing models are refreshed daily.

Is Agent Deck compatible with Windows without WSL?

No, Agent Deck requires WSL on Windows for full functionality, as specified in the README. It works on macOS, Linux, and Windows via WSL, with WSL2 recommended for features like MCP socket pooling and Docker sandboxing.

What happens if Docker isn't installed for sandboxing?

Docker sandbox features won't work, but other functionalities like session management, cost tracking, and MCP pooling remain available. You can disable sandboxing in the config or avoid Docker-specific commands, though isolation and security benefits are lost.

Open-Awesome

agent-deck

Q: How to set up MCP pooling in Agent Deck?

Enable MCP socket pooling by setting `pool_all = true` in the ~/.agent-deck/config.toml file. This shares MCP processes across sessions via Unix sockets, reducing memory usage by 85-90% with auto-recovery from crashes in about 3 seconds, as detailed in the configuration reference.

MITGov1.10.10

A terminal session manager that provides a unified TUI for managing multiple AI coding agents like Claude, Gemini, OpenCode, and Codex.

Visit Website GitHub

585 stars95 forks0 contributors

What is agent-deck?

Agent Deck is a terminal session manager specifically designed for AI coding agents. It provides a unified TUI to manage, monitor, and organize multiple agent sessions from tools like Claude Code, Gemini, OpenCode, and Codex in one place. It solves the problem of messy terminal sprawl and lack of visibility when running multiple AI agents concurrently across different projects.

Target Audience

Developers and engineers who regularly use multiple AI coding assistants (like Claude Code, Gemini CLI) across several projects and need to manage those sessions efficiently from the terminal.

Value Proposition

Developers choose Agent Deck because it offers AI-aware session management with features like smart status detection, session forking with context inheritance, MCP and skills management, and cost tracking—all in a single, organized TUI that integrates with existing tools without interference.

Overview

Terminal session manager for AI coding agents. One TUI for Claude, Gemini, OpenCode, Codex, and more.

Use Cases

Best For

Managing multiple Claude Code sessions across different projects simultaneously
Organizing AI agent sessions with Git worktrees to avoid repository conflicts
Reducing MCP server memory footprint when running many agent sessions
Tracking token usage and costs across all AI coding agent activities
Running AI agents in isolated Docker containers for security and sandboxing
Orchestrating agent sessions with persistent conductors for automated monitoring

Not Ideal For

Users who exclusively run a single AI agent session without need for centralized management
Teams relying on GUI-based AI tools or full IDE integrations without terminal access
Environments where Docker is unavailable or restricted, limiting sandboxing features
Projects requiring real-time multi-user collaboration beyond session monitoring

Pros & Cons

Pros

Memory-Efficient MCP Pooling

MCP socket pooling shares processes across sessions via Unix sockets, reducing memory usage by 85-90% with auto-recovery from crashes in ~3 seconds, as documented in the config reference.

Isolated Git Worktrees

Enables multiple agents to work on the same repository in isolated branches without conflicts, using Git worktrees with configurable locations and setup scripts for ignored files.

Docker Sandbox Security

Runs sessions in isolated Docker containers with bind-mounted project directories, protecting the system while allowing code access, and includes auto-cleanup and shell access options.

Comprehensive Cost Tracking

Provides real-time token usage and cost tracking with TUI and web dashboards, supporting 13 models and configurable budget limits, though some integrations are untested.

Conductor Orchestration

Persistent agent sessions monitor and orchestrate other sessions with optional Telegram/Slack integration for remote control and automated response escalation.

Cons

Untested Feature Reliability

Cost tracking for tools like Gemini and Codex is noted as untested in the README, which may lead to inconsistent or unreliable data for users of those tools.

Dependency-Heavy Architecture

Requires tmux and Docker for full functionality, adding setup complexity and potential compatibility issues in environments where these are not installed or preferred.

Configuration Overhead

Advanced features like MCP pooling, conductor setup, and custom tools involve detailed TOML configuration, which can be daunting and time-consuming to fine-tune.

Limited Tool Integration Depth

While multi-tool support exists, full features like forking and MCP management are primarily optimized for Claude Code, with other tools like Cursor having basic status detection only.

Frequently Asked Questions

Related Projects

InkOS

Story Creation AI Agent for novel, scripts, translation, interactive games, and IP content

Manage multiple Claude Code, OpenCode agents from either TUI or Web for easy access on mobile. Also supports Mistral Vibe, Codex CLI, Gemini CLI, Pi.dev, Copilot CLI, Factory Droid Coding.

Stars2,868

Forks277

Last commit9 hours ago

coi

Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.

Stars601

Forks50

Last commit12 hours ago

greywall

Container-free, deny-by-default sandbox for AI coding agents. Kernel-enforced filesystem, network, and syscall isolation for Linux and macOS

Stars275

Forks33

Last commit1 month ago

Community-curated · Updated weekly · 100% open source

Found a gem we're missing?

Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.

Submit a project Star on GitHub