robzilla1738/harness-terminal

Native macOS terminal with GPU rendering, durable sessions, coding-agent workflows, and SSH connectivity. Session host, daemon, and CLI support Linux; optional MCP and AI summaries.

What it solves

Harness is a native macOS terminal emulator designed to keep coding sessions persistent and provide deep integration with AI coding agents. It solves the problem of losing terminal state when an app is closed and the difficulty of tracking multiple AI agents running in different tabs, which often requires manual checking for when an agent has stopped or needs user approval.

How it works

Harness uses a separate session host and daemon to manage shells and PTYs, allowing sessions to survive application quits and restarts. It features a GPU-accelerated rendering engine for high performance and supports remote connectivity via a headless daemon that can be run on Linux or macOS, driven by a harness-cli over SSH. For AI agents, it detects processes from tools like Claude Code, Codex, and Cursor, and uses optional hooks to surface notifications and alerts when an agent requires attention.

Who it’s for

Developers and power users on macOS who use AI coding agents and frequently work across multiple local and remote sessions that they want to persist across restarts.

Highlights

  • Agent-Awareness: Detects 37 coding CLIs and provides notifications and a dedicated shortcut (Cmd+Shift+U) to jump directly to agents waiting for user input.
  • Session Persistence: Sessions are managed by a daemon, allowing them to survive app quits and can be attached to from the CLI or remotely.
  • GPU Rendering: Custom GPU engine with support for sRGB and Display-P3 color, ligatures, and inline images (Sixel/Kitty/iTerm2).
  • Remote & Headless: Ability to run a daemon on a remote Linux/macOS box and manage it via harness-cli over SSH.
  • Extensive Automation: A powerful CLI (harness-cli) for sending keys, capturing panes, and managing layouts programmatically.
  • Customization: Over 500 bundled themes and multiple experience modes ranging from a plain terminal to a full agent workspace.

Related

  • Project
  • Project
  • Project
  • Project