awesome-terminals

cdleon's catalogue of terminal emulators, shells, command-line tools, dotfiles and guides — 403 entries, 2,925★ / 159 forks / 19 open items / 33 watchers, 267 commits, created 2018-10-07, last pushed 2026-07-20. Counted from the README rather than taken from the summary. No license file at all, which is unusual and consequential for a directory (see below).

The spoke's third catalogue, after openalternative and best-of-lists, and the first that is not a map of substitutes.

What it holds

129 terminal emulators, grouped by where they run: Linux 48, macOS 37, Windows 26, iOS 7, Web 6, Android 4, HarmonyOS 1. The names a reader would expect are there — Alacritty, Kitty, WezTerm, Ghostty, Contour, foot, Rio, Tabby, Hyper, iTerm2, Windows Terminal — with one-line descriptions carrying the feature that distinguishes each ("GPU-accelerated", "Wayland", "modal").

253 command-line tools, grouped by implementation language: Go 55, Rust 48, Python 39, Node 28, C 23, C++ 10, Ruby 6, Haskell 3, Swift 1, plus shell-script sections. Then 8 shells, 13 dotfile repositories, and a short guides section.

Two axes, and neither one is substitution

This is the observation worth keeping, and it lands directly on the spoke's own argument.

open-source-alternative holds that the substitution frame — software defined by the proprietary product it replaces — is what sells this category and also what distorts it. openalternative is that frame made into a directory, ranked by GitHub stars and nothing else, and the spoke recorded that two of its own founding sources are invisible to it: proton-wine-andrevto replaces no proprietary product, and omniwm is a window manager for a platform that ships none.

awesome-terminals sorts on where a thing runs and what it was written in. Neither axis asks what anything replaces, so nothing here can fall through the gap those two sources fell through. That is not a virtue the maintainer claims — it is a consequence of cataloguing a software category rather than a market position, and it is the first evidence in this corpus that the choice of axis, not the choice of metric, is what decides what a directory can see.

The price is paid immediately: this is not a FOSS directory. iTerm2 sits beside Alacritty, and Cathode is a dead commercial product. A category axis admits everything in the category, licence irrelevant. So the list is genuinely useful to this spoke and is not evidence about open source specifically — a distinction worth holding, since the other two catalogues here are selected for it.

What it can tell you, against the other two

The deprecation tags are the interesting column. best-of-lists derives inactivity automatically from repository metadata, which is why 17 of its 22 lists carry the marker — the machinery documents its own decay. Here a human wrote [DEPRECATED] thirteen times, on eDEX-UI, Darktile, Notty, Upterm, Cathode and others, which is a judgement rather than a threshold. It cannot be applied consistently and it can say things a commit-date heuristic cannot, such as that a project's author declared it finished. Neither approach is better; they fail differently, and this corpus now holds both.

No ranking at all is the other break from the pattern. project-quality-score exists because openalternative ranks by stars and the spoke wanted to know what a better metric would change. This list declines the question — a reader gets a name, a link and a sentence, and does their own comparing. For a category where the choice is largely about feel and platform, that may be the honest answer.

The accident in the tools section

Sorting 253 CLI tools by implementation language produces a distribution nobody set out to publish: Go 55, Rust 48, Python 39, Node 28, C 23. Two languages that did not exist when most terminal tooling was written account for more than 40% of the list, and C — the language the terminal was built in — is fifth.

It is a convenience sample with no inclusion criteria, drawn from one maintainer's attention, so it is evidence of nothing in particular about language adoption. It is still the densest fact in the file, and it is a by-product of an organizing decision rather than a claim the list makes.

The missing licence

403 curated entries and no LICENSE file. Both sibling catalogues here are CC0-1.0, which matters because a directory's value is its data — reusing, forking or building on this list has no stated permission behind it. That is a sharper problem for a catalogue than for a tool, where the code being unlicensed at least leaves the idea free.

Standing and limits

T3 — a maintainer's curated list with no inclusion criteria, no dates on entries, no metric, and nothing verified about any tool it names. It is good evidence of its own taxonomy and audience and weak evidence about anything in it, which is the same reading the spoke gives openalternative. The counts above are a 2026-08-17 snapshot; a list pushed a month ago is maintained but not busy.

Connections. tuios arrived here the same day and is not in the list — a multiplexer rather than an emulator, which is exactly the boundary this catalogue draws by putting emulators under platforms and everything else under languages. And the spoke's environment-you-compute-in register (anysoftkeyboard, omniwm, termux, proot-distro, tuios) now has a map of one of its corners, which none of the others do.

Related

openalternative · best-of-lists · open-source-alternative · project-quality-score · tuios · omniwm · termux · foss-application · synthesis