Harness Terminal

Harness Terminal feels like the terminal Apple might build if they cared about developers using AI agents. It's a single Swift binary with no package dependencies—that's rare. The GPU renderer is smooth, ligatures and inline images just work, and the multiplexer is solid. We'd reach for this when we're deep in sessions with Claude Code or Codex and don't want to manually check if an agent is still running. The agent awareness is the killer feature: it watches the process tree and marks panes automatically, plus you can add hooks for notifications. Where it bites: macOS-only and Apple Silicon only, so no Linux or Intel Mac love. There's no plugin system—you get what's built in, which is fine for the focused use case but limiting if you want to script fancy integrations. It's also relatively new (v1.0 only four months old), so the community is small and the documentation is still maturing. Compared to iTerm2 + tmux: Harness replaces both with one app. iTerm2 has years of polish and a massive plugin ecosystem, but Harness's agent awareness is unique. If you don't use AI coding assistants, the agent features are wasted, and iTerm2/tmux might serve you better. For developers who do use agents, Harness is genuinely useful. In practice, the workspace sidebar is a big win—active sessions stay visible instead of buried. The CLI (harness-cli) lets you script layouts, and you can attach to remote sessions over SSH. The background daemon (HarnessDaemon) handles persistence even when the window is closed. It's free, open source (MIT), and the codebase is clean enough to fork if needed. We'd recommend it for macOS developers who want a modern, agent-aware terminal and don't mind being on an Apple Silicon-only island.