Quil: reboot-proof terminal multiplexer resumes AI sessions

Quil is a terminal multiplexer—a tool that lets you split your terminal into multiple panes and tabs—built from the ground up for developers who work with AI assistants. Unlike tmux (the Unix standard), Quil persists your entire workspace across reboots: when you power off your machine and power it back on, typing `quil` restores every tab, pane, working directory, layout split, and AI conversation to exactly where you left it, completing the restoration in roughly 30 seconds. This persistence works because Quil is project-aware and conversation-aware—it saves not just the terminal state but the context of each pane, including the session ID of any AI conversation running there.

The core workflow revolves around typed panes. Instead of generic shells, Quil offers built-in pane types for the tools developers run daily: Terminal (your system shell with live working-directory tracking), Claude Code and OpenCode (AI coding sessions that resume by session ID), lazygit (a Git TUI), k9s (a Kubernetes cluster manager), lazysql (a database client), SSH (persistent remote sessions), and Stripe CLI (for webhook tunnels). Each pane type knows how to restore itself; for example, an SSH pane re-runs the same host, port, and forwards. Users can also define custom pane types in TOML. New panes are created with Ctrl+N, tabs with Ctrl+T, and the command palette (Alt+Shift+P) fuzzy-finds every pane and tab—navigation grouped at the top, all available actions shown below with their keybindings.

The most significant feature for AI-native teams is MCP (Model Context Protocol) integration. By adding Quil to an AI client's MCP config (Claude Desktop, Claude Code, Cursor, or VS Code Copilot), developers give the AI assistant access to terminal actions: list_panes, read_pane_output, send_to_pane, watch_notifications, screenshot_pane, and 12 additional tools. This means an AI assistant can autonomously read the output of a failed build in one pane, react to the error, and send a fix command to another pane—all without the developer copying and pasting. For teams with high session counts, this eliminates a major friction point.

Recovery is built in. If a pane hangs, `quil restart` escalates through stop → fresh start → tabs restored from the last snapshot, or Alt+R restarts a single stuck pane in place with its AI session intact. Installation is one-line on Linux/macOS (`curl -sSfL https://raw.githubusercontent.com/artyomsv/quil/master/scripts/install.sh | sh`), downloadable on Windows, or via `go install` for Go users. The tool is licensed MIT and comes with comprehensive documentation covering installation, quick-start, keybindings, MCP setup, custom plugins, architecture (24 ADRs), and a roadmap.

Quil addresses a specific friction point in modern development: as teams adopt AI-native workflows (pairing with Claude, OpenAI, or other LLM-powered coding tools), they often run multiple long-lived sessions—an AI coding window, a build watcher, a webhook tunnel, an SSH connection—across many projects. Traditional terminal multiplexers like tmux can preserve shell state but do not understand AI conversations or provide a structured way for agents to interact with the terminal. Quil changes that by introducing typed panes (which know their purpose and restore strategy), session-aware resumption (so an AI conversation picks up exactly where it left off), and MCP integration (so an AI assistant can autonomously read pane output and send commands). For developers juggling 5–10 sessions per project, this means fewer manual context switches and less copy-pasting of error logs or command output to the AI—the AI can see and act on your terminal directly. The tool's availability across Linux, macOS, and Windows, combined with one-line install and a command palette, lowers the barrier to adoption.