keybr.com: self-hosting the adaptive typing tutor
The smartest way to learn touch typing and improve your typing speed.
At a glance
- What is it?
- keybr.com is an open source touch typing tutor that tracks every keystroke and generates lessons around your weakest keys. The repository is a TypeScript monorepo you can run locally or in Docker, and the AGPL-3.0 licence shapes who can reasonably fork it.
- Who is it for?
- Adopt keybr.com if you want a typing tutor whose lesson generator adapts to per-key statistics, and you are comfortable running a Node monorepo with SQLite behind it. Do not adopt it if you need a hosted service you never operate, or if you intend to fold the code into a closed product, because AGPL-3.0 applies.
- Can I use it commercially?
- Yes, with strict conditions. AGPL-3.0 is a network copyleft licence: if people use a modified version over a network, for example as a hosted service, you must offer them its source code under the same licence.
- Is it still maintained?
- Yes. The repository last received commits 48 days ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 13, 2026, and from our analysis. They are not legal advice.
Star history
- aradzie/keybr.com4,852 ★ · +119
| Repository | 2026-09-23 | 2026-11-13 |
|---|---|---|
| aradzie/keybr.com | 4.7k | 4.9k |
119 Stars in the last 51 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
What keybr.com actually does that a plain typing test does not
A typing test shows you a paragraph and reports words per minute. keybr.com does something different: the README states that it tracks every single keystroke and computes statistics for each individual key, then automatically generates lessons that focus on your weakest keys. That is the whole product in one sentence. The lesson text is not a fixed corpus you grind through; it is generated from your own error and timing history, key by key.
The progression model is the second half of the idea. According to the README, the tutor starts with a small set of the most frequent letters in your language, and more letters are added once you reach the target speed with the current ones. So the difficulty curve is gated on measured performance rather than on a lesson number. The README also says it can predict how many more lessons you will need to reach your target speed, and that a profile page graphs your progress.
Who is this for? Someone who has decided to learn touch typing properly and wants the practice material to react to their specific weak fingers, rather than practising on whatever text a static tutor ships with. It is a poor fit for someone who wants a quick one-off speed measurement and nothing else.
Per-key statistics, generated lessons, and the monorepo that delivers them
The repository is a TypeScript monorepo. package.json declares workspaces for packages/* and scripts, and the build runs through lage (the package.json scripts call lage compile, lage test and lage clean). The web bundle is built by webpack: the build script is env NODE_ENV=production webpack, and there is a watch script for development. The server entry point is root/index.js, started with node --enable-source-maps.
On the server side the dependency list is dominated by the @fastr family of packages (@fastr/core, @fastr/controller, @fastr/invert, @fastr/middleware-body, @fastr/middleware-compress and others). That is the HTTP layer the application is assembled from. Persistence is configurable: .env.example sets DATABASE_CLIENT=sqlite and DATABASE_FILENAME=~/.local/state/keybr/database.sqlite, with DATA_DIR=~/.local/state/keybr holding state. So the default deployment is a single Node process with a SQLite file on disk, which is what makes the Docker path in the repository plausible for a small self-hosted instance.
The README does not document the wire format or the statistics algorithm itself, so anyone wanting to modify how lessons are generated will be reading packages/ rather than documentation.
Installing keybr.com locally and running a first lesson
The repository ships two paths. The Docker path is the shorter one. docker-compose.yaml defines a single service built from ./Dockerfile, publishing port 30044 on the host to port 3000 in the container, with two volume mounts: a data directory mounted at ~/.local/state/keybr, and a .env file mounted at /etc/keybr/env. Replace both /path/to/data_dir placeholders before starting.
services:
keybr:
build:
dockerfile: ./Dockerfile
volumes:
- /path/to/data_dir:~/.local/state/keybr
- /path/to/data_dir/.env:/etc/keybr/env
ports:
- 30044:3000
restart: unless-stoppedThe Dockerfile itself is a node:26 image that copies the repository, runs npm ci, then npm run compile && npm run build, exposes 3000, and starts with npm run start-docker. Note that start-docker first runs packages/devenv/lib/initdb.ts before starting the server, which is how the database gets created inside the container.
For a local checkout without Docker, copy the environment template and edit it. The commands below call scripts defined in package.json.
cp .env.example .env
npm ci
npm run compile
npm run startWith the defaults, APP_URL is http://localhost:3000/ and the database is a SQLite file under ~/.local/state/keybr. Once the server is up, open the app in a browser and start a lesson; the README describes the experience as showing you a piece of text and having you type it out, with the letter set expanding as you hit the target speed. The README points contributors at docs/getting_started.md for launching a local instance, and that file is the maintainers' own account of the setup, so treat it as the authority if anything above diverges.
Where keybr.com is the wrong tool or the wrong deployment
The licence is the first constraint. The README states the project is released under the GNU Affero General Public License v3.0, and package.json carries an SPDX-style license field of GPL-3. If you want to embed the lesson generator in a proprietary product, or run a modified version as a network service without publishing your changes, AGPL-3.0 is the wrong licence for that plan. This is a description of the licence the repository declares, not legal advice; check with a lawyer for your situation.
Second, this is a server application, not a static page. Even with SQLite, you are operating a Node process, a build pipeline (npm run compile && npm run build) and a data directory that must persist. If your goal is a typing tutor for a classroom with no server, the operational surface here is larger than the task requires.
Third, the README is honest that the interface looks simple and that the interesting behaviour is in the details, but it does not document the lesson-generation algorithm, the statistics model or a rollback procedure. There is no releases list, so there is no versioned upgrade story to lean on; you would be tracking the master branch. If you need documented, versioned upgrade paths, that absence matters more than any feature list.
How keybr.com differs from Monkeytype and other typing sites
Monkeytype is the natural comparison point, and the difference is in what the practice text is for. Monkeytype is a testing and practice site: you choose a mode (time, words, quote), the text is drawn from a corpus or your own input, and the result is a score. The text is the input to the measurement. keybr.com inverts that. The README describes lessons generated automatically to focus on your weakest keys, with the active letter set starting from the most frequent letters in your language and growing as you hit the target speed. The measurement is the input to the text.
That difference has practical consequences. With a corpus-based tester you can compare two runs fairly because the material is comparable. With generated lessons, the material changes as you improve, which is the point, but it also means a raw WPM number from keybr.com is measuring you against a moving target. The README does not claim the two are directly comparable, and they are not.
A second difference is deployment. Monkeytype is usable as a hosted site and you never think about its database. keybr.com is also hosted at keybr.com, but the repository is the whole application, SQLite included, which is what makes the self-hosted path in docker-compose.yaml meaningful. You are choosing between using someone else's instance and running your own.
Maintenance, upgrades and what the AGPL-3.0 licence means for a fork
The repository is not archived, and the last push was on 2026-09-28. That is a current codebase, not an abandoned one. The repository has no published GitHub releases, so the project appears to be consumed from the master branch rather than from tagged versions. Practically, that means an upgrade is a git pull plus npm ci, npm run compile and npm run build, followed by a restart. The Dockerfile encodes exactly that sequence, so rebuilding the image is the safest upgrade procedure: it re-runs npm ci and the compile and build steps in a clean layer.
The runtime cost is modest by the standards of web applications: one Node process, one SQLite file, and a data directory. The docker-compose.yaml mounts the data directory and the env file separately, which is the right shape for backups: copy the directory under DATA_DIR and you have the database. Nothing here documents a migration tool, so schema changes between commits are a real risk when you pull; test against a copy of the database file before pointing your instance at the live one.
On licence implications, AGPL-3.0 is a copyleft licence with a network clause. The README states the project is released under it. If you run a modified keybr.com as a service other people use, the licence's network provision is the part to read carefully. If you only run it unmodified for yourself, the obligation is much lighter.
Editorial conclusion
Adopt keybr.com if you want a typing tutor whose lesson generator adapts to per-key statistics, and you are comfortable running a Node monorepo with SQLite behind it. Do not adopt it if you need a hosted service you never operate, or if you intend to fold the code into a closed product, because AGPL-3.0 applies. Before committing, run npm run compile and npm run test in the checkout, confirm the DATA_DIR and DATABASE_FILENAME values in .env.example match your storage, and read docs/getting_started.md for the maintainers' own setup path.
Frequently asked questions
Is keybr.com free?
The source is released under the GNU Affero General Public License v3.0, so you can run your own instance from the repository at no licence cost.
Is keybr.com safe to use?
There is no security audit in the repository, so there is no basis for a safety claim. What can be said is that self-hosting removes the third-party site from the picture: the repository ships a Dockerfile and docker-compose.yaml, and .env.example keeps data in a local SQLite file under DATA_DIR.
What is keybr.com?
It is a touch typing tutor. The README describes it as showing you a piece of text to type out, while tracking every keystroke, computing per-key statistics, and generating lessons that focus on your weakest keys.
How do I use keybr.com?
On the hosted site you open it and start typing the displayed text; the README says the letter set starts small with the most frequent letters in your language and grows as you reach the target speed. To run it yourself, build from the Dockerfile or run npm ci, npm run compile and npm run start in a checkout.
Is keybr.com good for learning touch typing?
The design targets exactly that: per-key statistics drive lesson generation, and the README states the tutor predicts how many more lessons you need to reach your target speed. Whether it suits you depends on whether you want generated practice text instead of a fixed corpus.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/aradzie-keybr-com)Star history chart for your README
[](https://hysenlabs.com/projects/aradzie-keybr-com)Was this useful?