- asdf-compatible - rtx is compatible with asdf plugins and
.tool-versions
files. It can be used as a drop-in replacement. - Polyglot - compatible with any language, so no more figuring out how nvm, nodenv, pyenv, etc work individually—just use 1 tool.
- Fast - rtx is written in Rust and is very fast. 20x-200x faster than asdf.
- No shims - shims cause problems, they break
which
, and add overhead. By default, rtx does not use them—however you can if you want to. - Fuzzy matching and aliases - It's enough to just say you want "v20" of node, or the "lts" version. rtx will figure out the right version without you needing to specify an exact version.
- Arbitrary env vars - Set custom env vars when in a project directory like
NODE_ENV=production
orAWS_PROFILE=staging
.
The following shows using rtx to install different versions
of node.
Note that calling which node
gives us a real path to node, not a shim.
Install rtx on macOS (other methods here):
$ curl https://rtx.pub/rtx-latest-macos-arm64 > ~/bin/rtx
$ chmod +x ~/bin/rtx
$ rtx --version
rtx 2023.8.7
Hook rtx into your shell (pick the right one for your shell):
# note this assumes rtx is located at ~/bin/rtx
echo 'eval "$(~/bin/rtx activate bash)"' >> ~/.bashrc
echo 'eval "$(~/bin/rtx activate zsh)"' >> ~/.zshrc
echo '~/bin/rtx activate fish | source' >> ~/.config/fish/config.fish
Warning
If you use direnv with
layout python
or other logic that needs to reference rtx runtimes inside of an.envrc
, see the direnv section below.
Install a runtime and set it as the global default:
$ rtx use --global node@20
$ node -v
v20.0.0
Click to expand
- Features
- 30 Second Demo
- Quickstart
- About
- Installation
- Uninstalling
- Shebang
- Configuration
- Aliases
- Plugins
- Versioning
- Directories
- Templates
- [experimental] Config Environments
- IDE Integration
- Core Plugins
- FAQs
- I don't want to put a
.tool-versions
file into my project since git shows it as an untracked file. - What is the difference between "nodejs" and "node" (or "golang" and "go")?
- What does
rtx activate
do? rtx activate
doesn't work in~/.profile
,~/.bash_profile
,~/.zprofile
- rtx is failing or not working right
- Windows support?
- How do I use rtx with http proxies?
- How do the shorthand plugin names map to repositories?
- Does "node@20" mean the newest available version of node?
- How do I migrate from asdf?
- How compatible is rtx with asdf?
- rtx isn't working when calling from tmux or another shell initialization script
- How do I disable/force CLI color output?
- Is rtx secure?
- I don't want to put a
- Comparison to asdf
- Shims
- direnv
- Cache Behavior
- Commands
rtx activate [OPTIONS] [SHELL_TYPE]
rtx alias get <PLUGIN> <ALIAS>
rtx alias ls [OPTIONS]
rtx alias set <PLUGIN> <ALIAS> <VALUE>
rtx alias unset <PLUGIN> <ALIAS>
rtx bin-paths
rtx cache clear
rtx completion [SHELL]
rtx current [PLUGIN]
rtx deactivate
rtx direnv activate
rtx doctor
rtx env [OPTIONS] [TOOL@VERSION]...
rtx env-vars [OPTIONS] [ENV_VARS]...
rtx exec [OPTIONS] [TOOL@VERSION]... [-- <COMMAND>...]
rtx implode [OPTIONS]
rtx install [OPTIONS] [TOOL@VERSION]...
rtx latest [OPTIONS] <TOOL@VERSION>
rtx link [OPTIONS] <TOOL@VERSION> <PATH>
rtx ls [OPTIONS]
rtx ls-remote <TOOL@VERSION> [PREFIX]
rtx outdated [TOOL@VERSION]...
rtx plugins install [OPTIONS] [NAME] [GIT_URL]
rtx plugins link [OPTIONS] <NAME> [PATH]
rtx plugins ls [OPTIONS]
rtx plugins ls-remote [OPTIONS]
rtx plugins uninstall <PLUGIN>...
rtx plugins update [PLUGIN]...
rtx prune [OPTIONS] [PLUGINS]...
rtx reshim
rtx self-update
rtx settings get <KEY>
rtx settings ls
rtx settings set <KEY> <VALUE>
rtx settings unset <KEY>
rtx shell [OPTIONS] [TOOL@VERSION]...
rtx sync node <--brew|--nvm|--nodenv>
rtx sync python --pyenv
rtx trust [OPTIONS] [CONFIG_FILE]
rtx uninstall <TOOL@VERSION>...
rtx upgrade [TOOL@VERSION]...
rtx use [OPTIONS] [TOOL@VERSION]...
rtx version
rtx where <TOOL@VERSION>
rtx which [OPTIONS] <BIN_NAME>
New developer? Try reading the Beginner's Guide for a gentler introduction.
rtx is a tool for managing programming language and tool versions. For example, use this to install
a particular version of node.js and ruby for a project. Using rtx activate
, you can have your
shell automatically switch to the correct node and ruby versions when you cd
into the project's
directory1. Other projects on your machine can use a different set of versions.
rtx is inspired by asdf and uses asdf's vast plugin ecosystem under the hood. However, it is much faster than asdf and has a more friendly user experience. For more on how rtx compares to asdf, see below.
rtx can be configured in many ways. The most typical is by .rtx.toml
, but it's also compatible
with asdf .tool-versions
files. It can also use idiomatic version files like .node-version
and
.ruby-version
. See Configuration for more.
rtx hooks into your shell (with rtx activate zsh
) and sets the PATH
environment variable to point your shell to the correct runtime binaries. When you cd
into a
directory1 containing a .tool-versions
/.rtx.toml
file, rtx will automatically set the
appropriate tool versions in PATH
.
After activating, every time your prompt displays it will call rtx hook-env
to fetch new
environment variables.
This should be very fast. It exits early if the directory wasn't changed or .tool-versions
/.rtx.toml
files haven't been modified.
Unlike asdf which uses shim files to dynamically locate runtimes when they're called, rtx modifies
PATH
ahead of time so the runtimes are called directly. This is not only faster since it avoids
any overhead, but it also makes it so commands like which node
work as expected. This also
means there isn't any need to run asdf reshim
after installing new runtime binaries.
You should note that rtx does not directly install these tools. Instead, it leverages plugins to install runtimes. See plugins below.
rtx install [email protected] Install a specific version number
rtx install node@20 Install a fuzzy version number
rtx use node@20 Use node-20.x in current project
rtx use -g node@20 Use node-20.x as global default
rtx install node Install the current version specified in .tool-versions/.rtx.toml
rtx use node@latest Use latest node in current directory
rtx use -g node@system Use system node as global default
rtx x node@20 -- node app.js Run `node app.js` node-20.x on PATH
Installing rtx consists of two steps.
- Download the binary. This depends on the device and operating system you are running rtx in.
- Register a shell hook. This depends on the shell you are using. Read more about this step in the FAQ.
Note that it isn't necessary for rtx
to be on PATH
. If you run the activate script in your rc
file, rtx will automatically add itself to PATH
.
curl https://rtx.pub/install.sh | sh
If you want to verify the install script hasn't been tampered with:
gpg --keyserver hkps://keyserver.ubuntu.com --recv-keys 0x29DDE9E0
curl https://rtx.pub/install.sh.sig | gpg --decrypt > install.sh
# ensure the above is signed with the rtx release key
sh ./install.sh
or if you're allergic to | sh
:
curl https://rtx.pub/rtx-latest-macos-arm64 > /usr/local/bin/rtx
It doesn't matter where you put it. So use ~/bin
, /usr/local/bin
, ~/.local/share/rtx/bin/rtx
or whatever.
Supported architectures:
x64
arm64
Supported platforms:
macos
linux
If you need something else, compile it with cargo. Windows isn't currently supported.
brew install rtx
Alternatively, use the custom tap (which is updated immediately after a release)):
brew install jdxcode/tap/rtx
sudo port install rtx
Build from source with Cargo:
cargo install rtx-cli
Do it faster with cargo-binstall:
cargo install cargo-binstall
cargo binstall rtx-cli
Build from the latest commit in main:
cargo install rtx-cli --git https://github.com/jdxcode/rtx --branch main
rtx is available on npm as a precompiled binary. This isn't a node.js package—just distributed
via npm. This is useful for JS projects that want to setup rtx via package.json
or npx
.
npm install -g rtx-cli
Use npx if you just want to test it out for a single command without fully installing:
npx rtx-cli exec [email protected] -- python some_script.py
Download the latest release from GitHub.
curl https://github.com/jdxcode/rtx/releases/download/v2023.8.7/rtx-v2023.8.7-linux-x64 > /usr/local/bin/rtx
chmod +x /usr/local/bin/rtx
For installation on Ubuntu/Debian:
sudo install -dm 755 /etc/apt/keyrings
wget -qO - https://rtx.pub/gpg-key.pub | gpg --dearmor | sudo tee /etc/apt/keyrings/rtx-archive-keyring.gpg 1> /dev/null
echo "deb [signed-by=/etc/apt/keyrings/rtx-archive-keyring.gpg arch=amd64] https://rtx.pub/deb stable main" | sudo tee /etc/apt/sources.list.d/rtx.list
sudo apt update
sudo apt install -y rtx
Warning
If you're on arm64 you'll need to run the following:
echo "deb [signed-by=/etc/apt/keyrings/rtx-archive-keyring.gpg arch=arm64] https://rtx.pub/deb stable main" | sudo tee /etc/apt/sources.list.d/rtx.list
For Fedora, CentOS, Amazon Linux, RHEL and other dnf-based distributions:
dnf install -y dnf-plugins-core
dnf config-manager --add-repo https://rtx.pub/rpm/rtx.repo
dnf install -y rtx
yum install -y yum-utils
yum-config-manager --add-repo https://rtx.pub/rpm/rtx.repo
yum install -y rtx
For Alpine Linux:
apk add rtx
rtx lives in the community repository.
For Arch Linux:
git clone https://aur.archlinux.org/rtx.git
cd rtx
makepkg -si
For the Nix package manager, at release 23.05 or later:
nix-env -iA rtx
You can also import the package directly using
rtx-flake.packages.${system}.rtx
. It supports all default Nix
systems.
echo 'eval "$(rtx activate bash)"' >> ~/.bashrc
echo 'rtx activate fish | source' >> ~/.config/fish/config.fish
do {
let rtxpath = ($nu.config-path | path dirname | path join "rtx.nu")
run-external rtx activate nu --redirect-stdout | save $rtxpath -f
$"\nsource "($rtxpath)"" | save $nu.config-path --append
}
Since .xsh
files are not compiled you may shave a bit off startup time by using a pure Python import: add the code below to, for example, ~/.config/xonsh/rtx.py
config file and import rtx
it in ~/.config/xonsh/rc.xsh
:
from pathlib import Path
from xonsh.built_ins import XSH
ctx = XSH.ctx
rtx_init = subprocess.run([Path('~/bin/rtx').expanduser(),'activate','xonsh'],capture_output=True,encoding="UTF-8").stdout
XSH.builtins.execx(rtx_init,'exec',ctx,filename='rtx')
Or continue to use rc.xsh
/.xonshrc
:
echo 'execx($(~/bin/rtx activate xonsh))' >> ~/.config/xonsh/rc.xsh # or ~/.xonshrc
Given that rtx
replaces both shell env $PATH
and OS environ PATH
, watch out that your configs don't have these two set differently (might throw os.environ['PATH'] = xonsh.built_ins.XSH.env.get_detyped('PATH')
at the end of a config to make sure they match)
Adding a new shell is not hard at all since very little shell code is in this project. See here for how the others are implemented. If your shell isn't currently supported I'd be happy to help you get yours integrated.
Use rtx implode
to uninstall rtx. This will remove the rtx binary and all of its data. Use
rtx implode --help
for more information.
Alternatively, manually remove the following directories to fully clean up:
~/.local/share/rtx
(can also beRTX_DATA_DIR
orXDG_DATA_HOME/rtx
)~/.config/rtx
(can also beRTX_CONFIG_DIR
orXDG_CONFIG_HOME/rtx
)- on Linux:
~/.cache/rtx
(can also beRTX_CACHE_DIR
orXDG_CACHE_HOME/rtx
) - on macOS:
~/Library/Caches/rtx
(can also beRTX_CACHE_DIR
)
You can specify a tool and its version in a shebang without needing to first
setup .tool-versions
/.rtx.toml
config:
#!/usr/bin/env -S rtx x node@20 -- node
// "env -S" allows multiple arguments in a shebang
console.log(`Running node: ${process.version}`);
This can also be useful in environments where rtx isn't activated (such as a non-interactive session).
.rtx.toml
is a new config file that replaces asdf-style .tool-versions
files with a file
that has lot more flexibility. It supports functionality that is not possible with .tool-versions
, such as:
- setting arbitrary env vars while inside the directory
- passing options to plugins like
virtualenv='.venv'
for rtx-python. - specifying custom plugin urls
Here is what an .rtx.toml
looks like:
[env]
# supports arbitrary env vars so rtx can be used like direnv/dotenv
NODE_ENV = 'production'
[tools]
# specify single or multiple versions
terraform = '1.0.0'
erlang = ['23.3', '24.0']
# supports everything you can do with .tool-versions currently
node = ['16', 'prefix:20', 'ref:master', 'path:~/.nodes/14']
# send arbitrary options to the plugin, passed as:
# RTX_TOOL_OPTS__VENV=.venv
python = {version='3.10', virtualenv='.venv'}
[plugins]
# specify a custom repo url
# note this will only be used if the plugin does not already exist
python = 'https://github.com/jdxcode/rtx-python'
[settings] # project-local settings
verbose = true
[alias.node] # project-local aliases
my_custom_node = '20'
.rtx.toml
files are hierarchical. The configuration in a file in the current directory will
override conflicting configuration in parent directories. For example, if ~/src/myproj/.rtx.toml
defines the following:
[tools]
node = '20'
python = '3.10'
And ~/src/myproj/backend/.rtx.toml
defines:
[tools]
node = '18'
ruby = '3.1'
Then when inside of ~/src/myproj/backend
, node
will be 18
, python
will be 3.10
, and ruby
will be 3.1
. You can check the active versions with rtx ls --current
.
You can also have environment specific config files like .rtx.production.toml
, see
Config Environments for more details.
The [env]
section of .rtx.toml allows setting arbitrary environment variables.
These can be simple key/value entries like this:
[env]
NODE_ENV = 'production'
PATH
is treated specially, it needs to be defined as an array in env_path
:
env_path = [
# adds an absolute path
"~/.local/share/bin",
# adds a path relative to the .rtx.toml, not PWD
"./node_modules/.bin",
]
Note: env_path
is a top-level key, it does not go inside of [env]
.
Environment variable values can be templates, see Templates for details.
[env]
LD_LIBRARY_PATH = "/some/path:{{env.LD_LIBRARY_PATH}}"
env_file
can be used to specify a dotenv file to load:
env_file = '.env'
Note: env_file
goes at the top of the file, above [env]
.
[env]
NODE_ENV = false # unset a previously set NODE_ENV
rtx supports "legacy version files" just like asdf. They're language-specific files like .node-version
and .python-version
. These are ideal for setting the runtime version of a project without forcing
other developers to use a specific tool like rtx/asdf.
They support aliases, which means you can have an .nvmrc
file with lts/hydrogen
and it will work
in rtx and nvm. Here are some of the supported legacy version files:
Plugin | "Legacy" (Idiomatic) Files |
---|---|
crystal | .crystal-version |
elixir | .exenv-version |
go | .go-version , go.mod |
java | .java-version , .sdkmanrc |
node | .nvmrc , .node-version |
python | .python-version |
ruby | .ruby-version , Gemfile |
terraform | .terraform-version , .packer-version , main.tf |
yarn | .yarnrc |
In rtx these are enabled by default. You can disable them with rtx settings set legacy_version_file false
.
There is a performance cost to having these when they're parsed as it's performed by the plugin in
bin/parse-version-file
. However these are cached so it's not a huge deal.
You may not even notice.
Note
asdf calls these "legacy version files" so we do too. I think this is a bad name since it implies that they shouldn't be used—which is definitely not the case IMO. I prefer the term "idiomatic" version files since they're version files not specific to asdf/rtx and can be used by other tools. (
.nvmrc
being a notable exception, which is tied to a specific tool.)
The .tool-versions
file is asdf's config file and it can be used in rtx just like .rtx.toml
.
It isn't as flexible so it's recommended to use .rtx.toml
instead. It can be useful if you
already have a lot of .tool-versions
files or work on a team that uses asdf.
Here is an example with all the supported syntax:
node 20.0.0 # comments are allowed
ruby 3 # can be fuzzy version
shellcheck latest # also supports "latest"
jq 1.6
erlang ref:master # compile from vcs ref
go prefix:1.19 # uses the latest 1.19.x version—needed in case "1.19" is an exact match
shfmt path:./shfmt # use a custom runtime
node lts # use lts version of node (not supported by all plugins)
node sub-2:lts # install 2 versions behind the latest lts (e.g.: 18 if lts is 20)
python sub-0.1:latest # install python-3.10 if the latest is 3.11
See the asdf docs for more info on this file format.
Both .rtx.toml
and .tool-versions
support "scopes" which modify the behavior of the version:
ref:<SHA>
- compile from a vcs (usually git) refprefix:<PREFIX>
- use the latest version that matches the prefix. Useful for Go since1.20
would only match1.20
exactly butprefix:1.20
will match1.20.1
and1.20.2
etc.path:<PATH>
- use a custom compiled version at the given path. One use-case is to re-use Homebrew tools (e.g.:path:/opt/homebrew/opt/node@20
).sub-<PARTIAL_VERSION>:<ORIG_VERSION>
- subtracts PARTIAL_VERSION from ORIG_VERSION. This can be used to express something like "2 versions behind lts" such assub-2:lts
. Or 1 minor version behind the latest version:sub-0.1:latest
.
rtx can be configured in ~/.config/rtx/config.toml
. It's like local .rtx.toml
files except that
it is used for all directories.
[tools]
# global tool versions go here
# you can set these with `rtx use -g`
node = 'lts'
python = ['3.10', '3.11']
[settings]
# plugins can read the versions files used by other version managers (if enabled by the plugin)
# for example, .nvmrc in the case of node's nvm
legacy_version_file = true # enabled by default (unlike asdf)
legacy_version_file_disable_tools = ['python'] # disable for specific tools
# configure `rtx install` to always keep the downloaded archive
always_keep_download = false # deleted after install by default
always_keep_install = false # deleted on failure by default
# configure how frequently (in minutes) to fetch updated plugin repository changes
# this is updated whenever a new runtime is installed
# (note: this isn't currently implemented but there are plans to add it: https://github.com/jdxcode/rtx/issues/128)
plugin_autoupdate_last_check_duration = '1 week' # set to 0 to disable updates
# config files with these prefixes will be trusted by default
trusted_config_paths = [
'~/work/my-trusted-projects',
]
verbose = false # set to true to see full installation output, see `RTX_VERBOSE`
asdf_compat = false # set to true to ensure .tool-versions will be compatible with asdf, see `RTX_ASDF_COMPAT`
jobs = 4 # number of plugins or runtimes to install in parallel. The default is `4`.
raw = false # set to true to directly pipe plugins to stdin/stdout/stderr
yes = false # set to true to automatically answer yes to all prompts
shorthands_file = '~/.config/rtx/shorthands.toml' # path to the shorthands file, see `RTX_SHORTHANDS_FILE`
disable_default_shorthands = false # disable the default shorthands, see `RTX_DISABLE_DEFAULT_SHORTHANDS`
disable_tools = ['node'] # disable specific tools, generally used to turn off core tools
experimental = false # enable experimental features
log_level = 'debug' # log verbosity, see `RTX_LOG_LEVEL`
[alias.node]
my_custom_node = '20' # makes `rtx install node@my_custom_node` install node-20.x
# this can also be specified in a plugin (see below in "Aliases")
These settings can also be managed with rtx settings ls|get|set|unset
.
rtx can also be configured via environment variables. The following options are available:
This is the directory where rtx stores plugins and tool installs. The default location is ~/.local/share/rtx
.
This is the directory where rtx stores internal cache. The default location is ~/.cache/rtx
on
Linux and
~/Library/Caches/rtx
on macOS.
This is the path to the config file. The default is ~/.config/rtx/config.toml
.
(Or $XDG_CONFIG_HOME/config.toml
if that is set)
Set to something other than ".tool-versions" to have rtx look for .tool-versions
files but with
a different name.
Set to something other than .rtx.toml
to have rtx look for .rtx.toml
config files with a different name.
Enables environment-specific config files such as .rtx.development.toml
.
Use this for different env vars or different tool versions in
development/staging/production environments. See
Config Environments for more on how
to use this feature.
Set the version for a runtime. For example, RTX_NODE_VERSION=20
will use [email protected] regardless
of what is set in .tool-versions
/.rtx.toml
.
Plugins can read the versions files used by other version managers (if enabled by the plugin)
for example, .nvmrc
in the case of node's nvm. See legacy version files for more
information.
Set to "0" to disable legacy version file parsing.
Disable legacy version file parsing for specific tools. Separate with ,
.
Set to 1
to default to using .rtx.toml
in rtx local
instead of .tool-versions
for
configuration.
For now this is not used by rtx use
which will only use .rtx.toml
unless --path
is specified.
This is a list of paths that rtx will automatically mark as
trusted. They can be separated with :
.
These change the verbosity of rtx.
You can also use RTX_DEBUG=1
, RTX_TRACE=1
, and RTX_QUIET=1
as well as
--log-level=trace|debug|info|warn|error
.
Output logs to a file.
Same as RTX_LOG_LEVEL
but for the log file output level. This is useful if you want
to store the logs but not have them litter your display.
Set to "1" to always keep the downloaded archive. By default it is deleted after install.
Set to "1" to always keep the install directory. By default it is deleted on failure.
This shows the installation output during rtx install
and rtx plugin install
.
This should likely be merged so it behaves the same as RTX_DEBUG=1
and we don't have
2 configuration for the same thing, but for now it is its own config.
Only output .tool-versions
files in rtx local|global
which will be usable by asdf.
This disables rtx functionality that would otherwise make these files incompatible with asdf.
Set the number plugins or runtimes to install in parallel. The default is 4
.
Set to "1" to directly pipe plugin scripts to stdin/stdout/stderr. By default stdin is disabled because when installing a bunch of plugins in parallel you won't see the prompt. Use this if a plugin accepts input or otherwise does not seem to be installing correctly.
Sets RTX_JOBS=1
because only 1 plugin script can be executed at a time.
Use a custom file for the shorthand aliases. This is useful if you want to share plugins within an organization.
The file should be in this toml format:
elixir = "https://github.com/my-org/rtx-elixir.git"
node = "https://github.com/my-org/rtx-node.git"
Disables the shorthand aliases for installing plugins. You will have to specify full urls when
installing plugins, e.g.: rtx plugin install node https://github.com/asdf-vm/asdf-node.git
Disables the specified tools. Separate with ,
. Generally used for core plugins but works with
all.
This will automatically answer yes or no to prompts. This is useful for scripting.
Enables experimental features.
rtx supports aliasing the versions of runtimes. One use-case for this is to define aliases for LTS
versions of runtimes. For example, you may want to specify lts-hydrogen
as the version for [email protected]
so you can use set it with node lts-hydrogen
in .tool-versions
/.rtx.toml
.
User aliases can be created by adding an alias.<PLUGIN>
section to ~/.config/rtx/config.toml
:
[alias.node]
my_custom_20 = '20'
Plugins can also provide aliases via a bin/list-aliases
script. Here is an example showing node.js
versions:
#!/usr/bin/env bash
echo "lts-hydrogen 18"
echo "lts-gallium 16"
echo "lts-fermium 14"
Note:
Because this is rtx-specific functionality not currently used by asdf it isn't likely to be in any plugin currently, but plugin authors can add this script without impacting asdf users.
rtx uses asdf's plugin ecosystem under the hood. These plugins contain shell scripts like
bin/install
(for installing) and bin/list-all
(for listing all of the available versions).
See https://github.com/asdf-vm/asdf-plugins for the list of built-in plugins shorthands. See asdf's Create a Plugin for how to create your own or just learn more about how they work.
rtx has support for "plugin options" which is configuration specified in .rtx.toml
to change behavior
of plugins. One example of this is virtualenv on python runtimes:
[tools]
python = {version='3.11', virtualenv='.venv'}
This will be passed to all plugin scripts as RTX_TOOL_OPTS__VIRTUALENV=.venv
. The user can specify
any option and it will be passed to the plugin in that format.
Currently this only supports simple strings, but we can make it compatible with more complex types (arrays, tables) fairly easily if there is a need for it.
rtx uses Calver versioning (2023.6.1
).
Breaking changes will be few but when they do happen,
they will be communicated in the CLI with plenty of notice whenever possible.
Rather than have semver major releases to communicate change in large releases,
new functionality and changes can be opted-into with settings like experimental = true
.
This way plugin authors and users can
test out new functionality immediately without waiting for a major release.
The numbers in Calver (YYYY.MM.RELEASE) simply represent the date of the release—not compatibility or how many new features were added. Each release will be small and incremental.
The following are the directories that rtx uses. These are the default directories, see Configuration for information on changing the locations.
Tip
If you often find yourself using these directories (as I do), I suggest setting all of them to
~/.rtx
for easy access.
This directory stores the global configuration file ~/.config/rtx/config.toml
.
On macOS this is ~/Library/Caches/rtx
.
Stores internal cache that rtx uses for things like the list of all available versions of a plugin. See Cache Behavior for more information.
This is the main directory that rtx uses and is where plugins and tools are installed into.
It is nearly identical to ~/.asdf
in asdf, so much so that you may be able to get by
symlinking these together and using asdf and rtx simultaneously. (Supporting this isn't a
project goal, however).
This is where plugins may optionally cache downloaded assets such as tarballs. Use the
always_keep_downloads
setting to prevent rtx from removing files from here.
rtx installs plugins to this directory when running rtx plugins install
. If you are working on a
plugin, I suggest
symlinking it manually by running:
ln -s ~/src/rtx-my-tool ~/.local/share/rtx/plugins/my-tool
This is where tools are installed to when running rtx install
. For example, rtx install [email protected]
will install to ~/.local/share/rtx/installs/node/20.0.0
This will also create other symlinks to this directory for version prefixes ("20" and "20.15") and matching aliases ("lts", "latest"). For example:
20 -> ./20.15.0
20.15 -> ./20.15.0
latest -> ./20.15.0
lts -> ./20.15.0
This is where rtx places shims. Generally these are used for IDE integration or if rtx activate
does not work for some reason.
Warning
This functionality is experimental and may change in the future.
Templates are used in the following locations:
.tool-versions
files.rtx.toml
files for most configuration- (Submit a ticket if you want to see it used elsewhere!)
The following context objects are available inside templates:
env: HashMap<String, String>
– current environment variablesconfig_root: PathBuf
– directory containing the.rtx.toml
file
As well as these functions:
exec(command: &str) -> String
– execute a command and return the output
Templates are parsed with tera—which is quite powerful. For example, this snippet will get the directory name of the project:
[env]
PROJECT_NAME = "{{config_root | split(pat='/') | last}}"
Here's another using exec()
:
[aliases]
current = "{{exec(command='node --version')}}"
It's possible to have separate .rtx.toml
files in the same directory for different
environments like development
and production
. To enable, set
experimental = true
in ~/.config/rtx/config.toml
, then set RTX_ENV
to an environment like
development
or production
. rtx will then look for a .rtx.{RTX_ENV}.toml
file in the current directory.
rtx will also look for "local" files like .rtx.local.toml
and .rtx.{RTX_ENV}.local.toml
in
the current directory. These are intended to not be committed to version control.
(Add rtx.*.local.toml
to your .gitignore
file.)
The priority of these files goes in this order (bottom overrides top):
.rtx.toml
.rtx.local.toml
.rtx.{RTX_ENV}.toml
.rtx.{RTX_ENV}.local.toml
Use rtx doctor
to see which files are being used.
Note that currently modifying RTX_DEFAULT_CONFIG_FILENAME
to something other than .rtx.toml
will not work with this feature. For now, it will disable it entirely. This may change in the
future.
IDEs work better with shims than they do environment variable modifications. The simplest way is to add the rtx shim directory to PATH.
For IntelliJ and VSCode—and likely others, you can modify ~/.zprofile
with the following:
export PATH="$HOME/.local/share/rtx/shims:$PATH"
This won't work for all of rtx's functionality. For example, arbitrary env vars in [env]
will only be set
if a shim is executed. For this we need tighter integration with the IDE and a custom plugin. If you feel
ambitious, take a look at existing direnv extensions for your IDE and see if you can modify it to work for rtx.
Direnv and rtx work similarly and there should be a direnv extension that can be used as a starting point.
Alternatively, you may be able to get tighter integration with a direnv extension and using the
use_rtx
direnv function.
rtx comes with some plugins built into the CLI written in Rust. These are new and will improve over
time. They can be easily overridden by installing a plugin with the same name, e.g.: rtx plugin install python
.
You can see the core plugins with rtx plugin ls --core
.
You can make git ignore these files in 3 different ways:
- Adding
.tool-versions
to project's.gitignore
file. This has the downside that you need to commit the change to the ignore file. - Adding
.tool-versions
to project's.git/info/exclude
. This file is local to your project so there is no need to commit it. - Adding
.tool-versions
to global gitignore (core.excludesFile
). This will cause git to ignore.tool-versions
files in all projects. You can explicitly add one to a project if needed withgit add --force .tool-versions
.
These are aliased. For example, rtx use [email protected]
is the same as rtx install [email protected]
. This
means it is not possible to have these be different plugins.
This is for convenience so you don't need to remember which one is the "official" name. However if something with the aliasing is acting up, submit a ticket or just stick to using "node" and "go". Under the hood, when rtx reads a config file or takes CLI input it will swap out "nodejs" and "golang".
While this change is rolling out, there is some migration code that will move installs/plugins from
the "nodejs" and "golang" directories to the new names. If this runs for you you'll see a message
but it should not run again unless there is some kind of problem. In this case, it's probably
easiest to just run rm -rf ~/.local/share/rtx/installs/{golang,nodejs} ~/.local/share/rtx/plugins/{golang,nodejs}
.
Once most users have migrated over this migration code will be removed.
It registers a shell hook to run rtx hook-env
every time the shell prompt is displayed.
rtx hook-env
checks the current env vars (most importantly PATH
but there are others like
GOROOT
or JAVA_HOME
for some tools) and adds/removes/updates the ones that have changed.
For example, if you cd
into a different directory that has java 18
instead of java 17
specified, just before the next prompt is displayed the shell runs: eval "$(rtx hook-env)"
which will execute something like this in the current shell session:
export JAVA_HOME=$HOME/.local/share/installs/java/18
export PATH=$HOME/.local/share/installs/java/18/bin:$PATH
In reality updating PATH
is a bit more complex than that because it also needs to remove java-17,
but you get the idea.
You may think that is excessive to run rtx hook-env
every time the prompt is displayed
and it should only run on cd
, however there are plenty of
situations where it needs to run without the directory changing, for example if .tool-versions
or
.rtx.toml
was just edited in the current shell.
Because it runs on prompt display, if you attempt to use rtx activate
in a
non-interactive session (like a bash script), it will never call rtx hook-env
and in effect will
never modify PATH because it never displays a prompt. For this type of setup, you can either call
rtx hook-env
manually every time you wish to update PATH, or use shims instead (preferred).
Or if you only need to use rtx for certain commands, just prefix the commands with
rtx x --
.
For example, rtx x -- npm test
or rtx x -- ./my_script.sh
.
rtx hook-env
will exit early in different situations if no changes have been made. This prevents
adding latency to your shell prompt every time you run a command. You can run rtx hook-env
yourself
to see what it outputs, however it is likely nothing if you're in a shell that has already been activated.
rtx activate
also creates a shell function (in most shells) called rtx
.
This is a trick that makes it possible for rtx shell
and rtx deactivate
to work without wrapping them in eval "$(rtx shell)"
.
rtx activate
should only be used in rc
files. These are the interactive ones used when
a real user is using the terminal. (As opposed to being executed by an IDE or something). The prompt
isn't displayed in non-interactive environments so PATH won't be modified.
For non-interactive setups, consider using shims instead which will route calls to the correct
directory by looking at PWD
every time they're executed. You can also call rtx exec
instead of
expecting things to be directly on PATH. You can also run rtx env
in a non-interactive shell, however that
will only setup the global tools. It won't modify the environment variables when entering into a
different project.
Also see the shebang example for a way to make scripts call rtx to get the runtime. That is another way to use rtx without activation.
First try setting RTX_DEBUG=1
or RTX_TRACE=1
and see if that gives you more information.
You can also set RTX_LOG_FILE_LEVEL=debug RTX_LOG_FILE=/path/to/logfile
to write logs to a file.
If something is happening with the activate hook, you can try disabling it and calling eval "$(rtx hook-env)"
manually.
It can also be helpful to use rtx env
which will just output environment variables that would be set.
Also consider using shims which can be more compatible.
If runtime installation isn't working right, try using the --raw
flag which will install things in
series and connect stdin/stdout/stderr directly to the terminal. If a plugin is trying to interact
with you for some reason this will make it work.
Of course check the version of rtx with rtx --version
and make sure it is the latest. Use rtx self-update
to update it. rtx cache clean
can be used to wipe the internal cache and rtx implode
can be used
to remove everything except config.
Before submitting a ticket, it's a good idea to test what you were doing with asdf. That way we can rule
out if the issue is with rtx or if it's with a particular plugin. For example, if rtx install python@latest
doesn't work, try running asdf install python latest
to see if it's an issue with asdf-python.
Lastly, there is rtx doctor
which will show diagnostic information and any warnings about issues
detected with your setup. If you submit a bug report, please include the output of rtx doctor
.
This is something we'd like to add! jdx#66
It's not a near-term goal and it would require plugin modifications, but it should be feasible.
Short answer: just set http_proxy
and https_proxy
environment variables. These should be lowercase.
rtx doesn't really do anything with http itself. The only exception to that is checking for new versions
and rtx self-update
. It uses git
to clone plugins and the plugins themselves generally will download
files with curl
or wget
.
However this is really up to the plugin. If you're having a proxy-related issue installing something you should post an issue on the plugin's repo.
e.g.: how does rtx plugin install node
know to fetch [https://github.com/rtx-plugins/rtx-nodejs]
(https://github.com/rtx-plugins/rtx-nodejs)?
asdf maintains an index of shorthands that rtx uses as a base. This is regularly updated every time that rtx has a release. This repository is stored directly into the codebase here. The bottom of that file contains modifications that rtx makes on top of asdf.
It depends on the command. Normally, for most commands and inside of config files, "node@20" will
point to the latest installed version of node-20.x. You can find this version by running
rtx latest --installed node@20
or by seeing what the ~/.local/share/rtx/installs/node/20
symlink
points to:
$ ls -l ~/.local/share/rtx/installs/node/20
[...] /home/jdxcode/.local/share/rtx/installs/node/20 -> node-v20.0.0-linux-x64
There are some exceptions to this, such as the following:
rtx install node@20
rtx latest node@20
rtx upgrade node@20
These will use the latest available version of node-20.x. This generally makes sense because you wouldn't want to install a version that is already installed.
First, just install rtx with rtx activate
like in the getting started guide and remove asdf from your
shell rc file.
Then you can just run rtx install
in a directory with an asdf .tool-versions
file and it will
install the runtimes. You could attempt to avoid this by copying the internal directory from asdf over
to rtx with cp -r ~/.asdf ~/.local/share/rtx
. That should work because they use the same structure,
however this isn't officially supported or regularly tested. Alternatively you can set RTX_DATA_DIR=~/.asdf
and see what happens.
rtx should be able to read/install any .tool-versions
file used by asdf. Any asdf plugin
should be usable in rtx. The commands in rtx are slightly
different, such as rtx install [email protected]
vs asdf install node 20.0.0
—this is done so
multiple tools can be specified at once. However, asdf-style syntax is still supported: (rtx install node 20.0.0
). This is the case for most commands, though the help for the command may
say that asdf-style syntax is supported.
When in doubt, just try asdf syntax and see if it works. If it doesn't open a ticket. It may not be possible to support every command identically, but we should attempt to make things as consistent as possible.
This isn't important for usability reasons so much as making it so plugins continue to work that call asdf commands.
If you need to switch to/from asdf or work in a project with asdf users, you can set
RTX_ASDF_COMPAT=1
. That prevents
rtx from writing .tool-versions
files that will not be
compatible with asdf. Also consider using .rtx.toml
instead which won't conflict with asdf setups.
rtx activate
will not update PATH until the shell prompt is displayed. So if you need to access a
tool provided by rtx before the prompt is displayed you must manually call hook-env
:
eval "$(rtx activate bash)"
eval "$(rtx hook-env)"
python --version # will work only after calling hook-env explicitly
For more information, see What does rtx activate
do?
rtx uses console.rs which honors the clicolors spec:
CLICOLOR != 0
: ANSI colors are supported and should be used when the program isn’t piped.CLICOLOR == 0
: Don’t output ANSI color escape codes.CLICOLOR_FORCE != 0
: ANSI colors should be enabled no matter what.
Not as much as it should be, though currently a bit more secure than asdf. Work will happen in this area as secure supply chains are incredibly important. See SECURITY.md for more information.
rtx is mostly a clone of asdf, but there are notable areas where improvements have been made.
asdf made (what I consider) a poor design decision to use shims that go between a call to a runtime
and the runtime itself. e.g.: when you call node
it will call an asdf shim file ~/.asdf/shims/node
,
which then calls asdf exec
, which then calls the correct version of node.
These shims have terrible performance, adding ~120ms to every runtime call. rtx does not use shims and instead
updates PATH
so that it doesn't have any overhead when simply calling binaries. These shims are the main reason that I wrote this. Note that in the demo gif at the top of this README
that rtx
isn't actually used when calling node -v
for this reason. The performance is
identical to running node without using rtx.
I don't think it's possible for asdf to fix these issues. The author of asdf did a great writeup of performance problems. asdf is written in bash which certainly makes it challenging to be performant, however I think the real problem is the shim design. I don't think it's possible to fix that without a complete rewrite.
rtx does call an internal command rtx hook-env
every time the directory has changed, but because
it's written in Rust, this is very quick—taking ~10ms on my machine. 4ms if there are no changes, 14ms if it's
a full reload.
tl;dr: asdf adds overhead (~120ms) when calling a runtime, rtx adds a small amount of overhead (~10ms) when the prompt loads.
asdf only helps manage runtime executables. However, some tools are managed via environment variables
(notably Java which switches via JAVA_HOME
). This isn't supported very well in asdf and requires
a separate shell extension just to manage.
However asdf plugins have a bin/exec-env
script that is used for exporting environment variables
like JAVA_HOME
. rtx simply exports
the environment variables from the bin/exec-env
script in the plugin but places them in the shell
for all commands. In asdf it only exports those commands when the shim is called. This means if you
call java
it will set JAVA_HOME
, but not if you call some Java tool like mvn
.
This means we're just using the existing plugin script but because rtx doesn't use shims it can be used for more things. It would be trivial to make a plugin that exports arbitrary environment variables like dotenv or direnv.
Some commands are the same in asdf but others have been changed. Everything that's possible
in asdf should be possible in rtx but may use slightly different syntax. rtx has more forgiving commands,
such as using fuzzy-matching, e.g.: rtx install node@20
. While in asdf you can run
asdf install node latest:20
, you can't use latest:20
in a .tool-versions
file or many other places.
In rtx
you can use fuzzy-matching everywhere.
asdf requires several steps to install a new runtime if the plugin isn't installed, e.g.:
asdf plugin add node
asdf install node latest:20
asdf local node latest:20
In rtx
this can all be done in a single step to set the local runtime version. If the plugin
and/or runtime needs to be installed it will prompt:
I've found asdf to be particularly rigid and difficult to learn. It also made strange decisions like
having asdf list all
but asdf latest --all
(why is one a flag and one a positional argument?).
rtx
makes heavy use of aliases so you don't need to remember if it's rtx plugin add node
or
rtx plugin install node
. If I can guess what you meant, then I'll try to get rtx to respond
in the right way.
That said, there are a lot of great things about asdf. It's the best multi-runtime manager out there and I've really been impressed with the plugin system. Most of the design decisions the authors made were very good. I really just have 2 complaints: the shims and the fact it's written in Bash.
Using rtx in CI/CD is a great way to synchronize tool versions for dev/build.
Use jdxcode/rtx-action
:
jobs:
lint:
runs-on: ubuntu-latest
steps:
- uses: jdxcode/rtx-action@v1
- run: node -v # will be the node version from `.rtx.toml`/`.tool-versions`
While the PATH design of rtx works great in most cases, there are some situations where shims are preferable. One example is when calling rtx binaries from an IDE.
To support this, rtx does have a shim dir that can be used. It's located at ~/.local/share/rtx/shims
.
$ rtx i [email protected]
$ rtx reshim # may be required if new shims need to be created
$ ~/.local/share/rtx/shims/node -v
v20.0.0
direnv and rtx both manage environment variables based on directory. Because they both analyze the current environment variables before and after their respective "hook" commands are run, they can sometimes conflict with each other.
If you have an issue, it's likely to do with the ordering of PATH. This means it would
really only be a problem if you were trying to manage the same tool with direnv and rtx. For example,
you may use layout python
in an .envrc
but also be maintaining a .tool-versions
file with python
in it as well.
A more typical usage of direnv would be to set some arbitrary environment variables, or add unrelated binaries to PATH. In these cases, rtx will not interfere with direnv.
If you do encounter issues with rtx activate
, or just want to use direnv in an alternate way,
this is a simpler setup that's less likely to cause issues—at the cost of functionality.
This may be required if you want to use direnv's layout python
with rtx. Otherwise there are
situations where rtx will override direnv's PATH. use rtx
ensures that direnv always has control.
To do this, first use rtx
to build a use_rtx
function that you can use in .envrc
files:
rtx direnv activate > ~/.config/direnv/lib/use_rtx.sh
Now in your .envrc
file add the following:
use rtx
direnv will now call rtx to export its environment variables. You'll need to make sure to add use_rtx
to all projects that use rtx (or use direnv's source_up
to load it from a subdirectory). You can also add use rtx
to ~/.config/direnv/direnvrc
.
Note that in this method direnv typically won't know to refresh .tool-versions
files
unless they're at the same level as a .envrc
file. You'll likely always want to have
a .envrc
file next to your .tool-versions
for this reason. To make this a little
easier to manage, I encourage not actually using .tool-versions
at all, and instead
setting environment variables entirely in .envrc
:
export RTX_NODE_VERSION=20.0.0
export RTX_PYTHON_VERSION=3.11
Of course if you use rtx activate
, then these steps won't have been necessary and you can use rtx
as if direnv was not used.
If you continue to struggle, you can also try using the shims method.
While making rtx compatible with direnv is, and will always be a major goal of this project, I also
want rtx to be capable of replacing direnv if needed. This is why rtx includes support for managing
env vars and virtualenv
for python using .rtx.toml
.
If you find you continue to need direnv, please open an issue and let me know what it is to see if
it's something rtx could support. rtx will never be as capable as direnv with a DSL like .envrc
,
but I think we can handle enough common use cases to make that unnecessary for most people.
rtx makes use of caching in many places in order to be efficient. The details about how long to keep cache for should eventually all be configurable. There may be gaps in the current behavior where things are hardcoded, but I'm happy to add more settings to cover whatever config is needed.
Below I explain the behavior it uses around caching. If you're seeing behavior where things don't appear to be updating, this is a good place to start.
Each plugin has a cache that's stored in ~/$RTX_CACHE_DIR/<PLUGIN>
. It stores
the list of versions available for that plugin (rtx ls-remote <PLUGIN>
), the legacy filenames (see below),
the list of aliases, the bin directories within each runtime installation, and the result of
running exec-env
after the runtime was installed.
Remote versions are updated daily by default. The file is zlib messagepack, if you want to view it you can run the following (requires msgpack-cli).
cat ~/$RTX_CACHE_DIR/node/remote_versions.msgpack.z | perl -e 'use Compress::Raw::Zlib;my $d=new Compress::Raw::Zlib::Inflate();my $o;undef $/;$d->inflate(<>,$o);print $o;' | msgpack-cli decode
Note that the caching of exec-env
may be problematic if the script isn't simply exporting
static values. The vast majority of exec-env
scripts only export static values, but if you're
working with a plugin that has a dynamic exec-env
submit
a ticket and we can try to figure out what to do.
Caching exec-env
massively improved the performance of rtx since it requires calling bash
every time rtx is initialized. Ideally, we can keep this
behavior.
Initializes rtx in the current shell
This should go into your shell's rc file.
Otherwise, it will only take effect in the current session.
(e.g. ~/.zshrc, ~/.bashrc)
This is only intended to be used in interactive sessions, not scripts.
rtx is only capable of updating PATH when the prompt is displayed to the user.
For non-interactive use-cases, use shims instead.
Typically this can be added with something like the following:
echo 'eval "$(rtx activate)"' >> ~/.zshrc
However, this requires that "rtx" is in your PATH. If it is not, you need to
specify the full path like this:
echo 'eval "$(/path/to/rtx activate)"' >> ~/.zshrc
Usage: activate [OPTIONS] [SHELL_TYPE]
Arguments:
[SHELL_TYPE]
Shell type to generate the script for
[possible values: bash, fish, nu, xonsh, zsh]
Options:
--status
Show "rtx: <PLUGIN>@<VERSION>" message when changing directories
Examples:
$ eval "$(rtx activate bash)"
$ eval "$(rtx activate zsh)"
$ rtx activate fish | source
$ execx($(rtx activate xonsh))
Show an alias for a plugin
This is the contents of an alias.<PLUGIN> entry in ~/.config/rtx/config.toml
Usage: alias get <PLUGIN> <ALIAS>
Arguments:
<PLUGIN>
The plugin to show the alias for
<ALIAS>
The alias to show
Examples:
$ rtx alias get node lts-hydrogen
20.0.0
List aliases
Shows the aliases that can be specified.
These can come from user config or from plugins in `bin/list-aliases`.
For user config, aliases are defined like the following in `~/.config/rtx/config.toml`:
[alias.node]
lts = "20.0.0"
Usage: alias ls [OPTIONS]
Options:
-p, --plugin <PLUGIN>
Show aliases for <PLUGIN>
Examples:
$ rtx aliases
node lts-hydrogen 20.0.0
Add/update an alias for a plugin
This modifies the contents of ~/.config/rtx/config.toml
Usage: alias set <PLUGIN> <ALIAS> <VALUE>
Arguments:
<PLUGIN>
The plugin to set the alias for
<ALIAS>
The alias to set
<VALUE>
The value to set the alias to
Examples:
$ rtx alias set node lts-hydrogen 18.0.0
Clears an alias for a plugin
This modifies the contents of ~/.config/rtx/config.toml
Usage: alias unset <PLUGIN> <ALIAS>
Arguments:
<PLUGIN>
The plugin to remove the alias from
<ALIAS>
The alias to remove
Examples:
$ rtx alias unset node lts-hydrogen
List all the active runtime bin paths
Usage: bin-paths
Deletes all cache files in rtx
Usage: cache clear
Generate shell completions
Usage: completion [SHELL]
Arguments:
[SHELL]
Shell type to generate completions for
[possible values: bash, elvish, fish, powershell, zsh]
Examples:
$ rtx completion bash > /etc/bash_completion.d/rtx
$ rtx completion zsh > /usr/local/share/zsh/site-functions/_rtx
$ rtx completion fish > ~/.config/fish/completions/rtx.fish
Shows current active and installed runtime versions
This is similar to `rtx ls --current`, but this only shows the runtime
and/or version. It's designed to fit into scripts more easily.
Usage: current [PLUGIN]
Arguments:
[PLUGIN]
Plugin to show versions of e.g.: ruby, node
Examples:
# outputs `.tool-versions` compatible format
$ rtx current
python 3.11.0 3.10.0
shfmt 3.6.0
shellcheck 0.9.0
node 20.0.0
$ rtx current node
20.0.0
# can output multiple versions
$ rtx current python
3.11.0 3.10.0
Disable rtx for current shell session
This can be used to temporarily disable rtx in a shell session.
Usage: deactivate
Examples:
$ rtx deactivate bash
$ rtx deactivate zsh
$ rtx deactivate fish
$ execx($(rtx deactivate xonsh))
Output direnv function to use rtx inside direnv
See https://github.com/jdxcode/rtx#direnv for more information
Because this generates the legacy files based on currently installed plugins,
you should run this command after installing new plugins. Otherwise
direnv may not know to update environment variables when legacy file versions change.
Usage: direnv activate
Examples:
$ rtx direnv activate > ~/.config/direnv/lib/use_rtx.sh
$ echo 'use rtx' > .envrc
$ direnv allow
Check rtx installation for possible problems.
Usage: doctor
Examples:
$ rtx doctor
[WARN] plugin node is not installed
Exports env vars to activate rtx a single time
Use this if you don't want to permanently install rtx. It's not necessary to
use this if you have `rtx activate` in your shell rc file.
Usage: env [OPTIONS] [TOOL@VERSION]...
Arguments:
[TOOL@VERSION]...
Tool(s) to use
Options:
-s, --shell <SHELL>
Shell type to generate environment variables for
[possible values: bash, fish, nu, xonsh, zsh]
--json
Output in JSON format
[short aliases: J]
Examples:
$ eval "$(rtx env -s bash)"
$ eval "$(rtx env -s zsh)"
$ rtx env -s fish | source
$ execx($(rtx env -s xonsh))
Manage environment variables
By default this command modifies ".rtx.toml" in the current directory.
You can specify the file name by either setting the RTX_DEFAULT_CONFIG_FILENAME environment variable, or by using the --file option.
Usage: env-vars [OPTIONS] [ENV_VARS]...
Arguments:
[ENV_VARS]...
Environment variable(s) to set
e.g.: NODE_ENV=production
Options:
--file <FILE>
The TOML file to update
Defaults to RTX_DEFAULT_CONFIG_FILENAME environment variable, or ".rtx.toml".
--remove <ENV_VAR>
Remove the environment variable from config file
Can be used multiple times.
Execute a command with tool(s) set
use this to avoid modifying the shell session or running ad-hoc commands with rtx tools set.
Tools will be loaded from .rtx.toml/.tool-versions, though they can be overridden with <RUNTIME> args
Note that only the plugin specified will be overridden, so if a `.tool-versions` file
includes "node 20" but you run `rtx exec [email protected]`; it will still load node@20.
The "--" separates runtimes from the commands to pass along to the subprocess.
Usage: exec [OPTIONS] [TOOL@VERSION]... [-- <COMMAND>...]
Arguments:
[TOOL@VERSION]...
Tool(s) to start e.g.: node@20 [email protected]
[COMMAND]...
Command string to execute (same as --command)
Options:
-c, --command <C>
Command string to execute
--cd <CD>
Change to this directory before executing the command
[short aliases: C]
Examples:
$ rtx exec node@20 -- node ./app.js # launch app.js using node-20.x
$ rtx x node@20 -- node ./app.js # shorter alias
# Specify command as a string:
$ rtx exec node@20 [email protected] --command "node -v && python -V"
# Run a command in a different directory:
$ rtx x -C /path/to/project node@20 -- node ./app.js
Removes rtx CLI and all related data
Skips config directory by default.
Usage: implode [OPTIONS]
Options:
--config
Also remove config directory
--dry-run
List directories that would be removed without actually removing them
Install a tool version
This will install a tool version to `~/.local/share/rtx/installs/<PLUGIN>/<VERSION>`
It won't be used simply by being installed, however.
For that, you must set up a `.rtx.toml`/`.tool-version` file manually or with `rtx use`.
Or you can call a tool version explicitly with `rtx exec <TOOL>@<VERSION> -- <COMMAND>`.
Tools will be installed in parallel. To disable, set `--jobs=1` or `RTX_JOBS=1`
Usage: install [OPTIONS] [TOOL@VERSION]...
Arguments:
[TOOL@VERSION]...
Tool(s) to install e.g.: node@20
Options:
-f, --force
Force reinstall even if already installed
-v, --verbose...
Show installation output
Examples:
$ rtx install [email protected] # install specific node version
$ rtx install node@20 # install fuzzy node version
$ rtx install node # install version specified in .tool-versions or .rtx.toml
$ rtx install # installs everything specified in .tool-versions or .rtx.toml
Gets the latest available version for a plugin
Usage: latest [OPTIONS] <TOOL@VERSION>
Arguments:
<TOOL@VERSION>
Tool to get the latest version of
Options:
-i, --installed
Show latest installed instead of available version
Examples:
$ rtx latest node@20 # get the latest version of node 20
20.0.0
$ rtx latest node # get the latest stable version of node
20.0.0
Symlinks a tool version into rtx
Use this for adding installs either custom compiled outside
rtx or built with a different tool.
Usage: link [OPTIONS] <TOOL@VERSION> <PATH>
Arguments:
<TOOL@VERSION>
Tool name and version to create a symlink for
<PATH>
The local path to the tool version
e.g.: ~/.nvm/versions/node/v20.0.0
Options:
-f, --force
Overwrite an existing tool version if it exists
Examples:
# build node-20.0.0 with node-build and link it into rtx
$ node-build 20.0.0 ~/.nodes/20.0.0
$ rtx link [email protected] ~/.nodes/20.0.0
# have rtx use the python version provided by Homebrew
$ brew install node
$ rtx link node@brew $(brew --prefix node)
$ rtx use node@brew
List installed and/or currently selected tool versions
Usage: ls [OPTIONS]
Options:
-p, --plugin <PLUGIN>
Only show tool versions from [PLUGIN]
-c, --current
Only show tool versions currently specified in a .tool-versions/.rtx.toml
-g, --global
Only show tool versions currently specified in a the global .tool-versions/.rtx.toml
-i, --installed
Only show tool versions that are installed Hides missing ones defined in .tool-versions/.rtx.toml but not yet installed
--json
Output in json format
[short aliases: J]
-m, --missing
Display missing tool versions
--prefix <PREFIX>
Display versions matching this prefix
Examples:
$ rtx ls
node 20.0.0 ~/src/myapp/.tool-versions latest
python 3.11.0 ~/.tool-versions 3.10
python 3.10.0
$ rtx ls --current
node 20.0.0 ~/src/myapp/.tool-versions 20
python 3.11.0 ~/.tool-versions 3.11.0
$ rtx ls --json
{
"node": [
{
"version": "20.0.0",
"install_path": "/Users/jdx/.rtx/installs/node/20.0.0",
"source": {
"type": ".rtx.toml",
"path": "/Users/jdx/.rtx.toml"
}
}
],
"python": [...]
}
List runtime versions available for install
note that the results are cached for 24 hours
run `rtx cache clean` to clear the cache and get fresh results
Usage: ls-remote <TOOL@VERSION> [PREFIX]
Arguments:
<TOOL@VERSION>
Plugin to get versions for
[PREFIX]
The version prefix to use when querying the latest version
same as the first argument after the "@"
Examples:
$ rtx ls-remote node
18.0.0
20.0.0
$ rtx ls-remote node@20
20.0.0
20.1.0
$ rtx ls-remote node 20
20.0.0
20.1.0
[experimental] Shows outdated tool versions
Usage: outdated [TOOL@VERSION]...
Arguments:
[TOOL@VERSION]...
Tool(s) to show outdated versions for
e.g.: node@20 [email protected]
If not specified, all tools in global and local configs will be shown
Examples:
$ rtx outdated
Plugin Requested Current Latest
python 3.11 3.11.0 3.11.1
node 20 20.0.0 20.1.0
$ rtx outdated node
Plugin Requested Current Latest
node 20 20.0.0 20.1.0
Install a plugin
note that rtx automatically can install plugins when you install a tool
e.g.: `rtx install node@20` will autoinstall the node plugin
This behavior can be modified in ~/.config/rtx/config.toml
Usage: plugins install [OPTIONS] [NAME] [GIT_URL]
Arguments:
[NAME]
The name of the plugin to install
e.g.: node, ruby
Can specify multiple plugins: `rtx plugins install node ruby python`
[GIT_URL]
The git url of the plugin
Options:
-f, --force
Reinstall even if plugin exists
-a, --all
Install all missing plugins
This will only install plugins that have matching shorthands.
i.e.: they don't need the full git repo url
-v, --verbose...
Show installation output
Examples:
# install the node via shorthand
$ rtx plugins install node
# install the node plugin using a specific git url
$ rtx plugins install node https://github.com/rtx-plugins/rtx-nodejs.git
# install the node plugin using the git url only
# (node is inferred from the url)
$ rtx plugins install https://github.com/rtx-plugins/rtx-nodejs.git
# install the node plugin using a specific ref
$ rtx plugins install node https://github.com/rtx-plugins/rtx-nodejs.git#v1.0.0
Symlinks a plugin into rtx
This is used for developing a plugin.
Usage: plugins link [OPTIONS] <NAME> [PATH]
Arguments:
<NAME>
The name of the plugin
e.g.: node, ruby
[PATH]
The local path to the plugin
e.g.: ./rtx-node
Options:
-f, --force
Overwrite existing plugin
Examples:
# essentially just `ln -s ./rtx-node ~/.local/share/rtx/plugins/node`
$ rtx plugins link node ./rtx-node
# infer plugin name as "node"
$ rtx plugins link ./rtx-node
List installed plugins
Can also show remotely available plugins to install.
Usage: plugins ls [OPTIONS]
Options:
-c, --core
The built-in plugins only
Normally these are not shown
-u, --urls
Show the git url for each plugin
e.g.: https://github.com/asdf-vm/asdf-node.git
--refs
Show the git refs for each plugin
e.g.: main 1234abc
Examples:
$ rtx plugins ls
node
ruby
$ rtx plugins ls --urls
node https://github.com/asdf-vm/asdf-node.git
ruby https://github.com/asdf-vm/asdf-ruby.git
List all available remote plugins
The full list is here: https://github.com/jdxcode/rtx/blob/main/src/default_shorthands.rs
Examples:
$ rtx plugins ls-remote
Usage: plugins ls-remote [OPTIONS]
Options:
-u, --urls
Show the git url for each plugin e.g.: https://github.com/rtx-plugins/rtx-nodejs.git
--only-names
Only show the name of each plugin by default it will show a "*" next to installed plugins
Removes a plugin
Usage: plugins uninstall <PLUGIN>...
Arguments:
<PLUGIN>...
Plugin(s) to remove
Examples:
$ rtx uninstall node
Updates a plugin to the latest version
note: this updates the plugin itself, not the runtime versions
Usage: plugins update [PLUGIN]...
Arguments:
[PLUGIN]...
Plugin(s) to update
Examples:
$ rtx plugins update # update all plugins
$ rtx plugins update node # update only node
$ rtx plugins update node#beta # specify a ref
Delete unused versions of tools
rtx tracks which config files have been used in ~/.local/share/rtx/tracked_config_files
Versions which are no longer the latest specified in any of those configs are deleted.
Versions installed only with environment variables (`RTX_<PLUGIN>_VERSION`) will be deleted,
as will versions only referenced on the command line (`rtx exec <PLUGIN>@<VERSION>`).
Usage: prune [OPTIONS] [PLUGINS]...
Arguments:
[PLUGINS]...
Prune only versions from these plugins
Options:
--dry-run
Do not actually delete anything
Examples:
$ rtx prune --dry-run
rm -rf ~/.local/share/rtx/versions/node/20.0.0
rm -rf ~/.local/share/rtx/versions/node/20.0.1
rebuilds the shim farm
This creates new shims in ~/.local/share/rtx/shims for CLIs that have been added.
rtx will try to do this automatically for commands like `npm i -g` but there are
other ways to install things (like using yarn or pnpm for node) that rtx does
not know about and so it will be necessary to call this explicitly.
If you think rtx should automatically call this for a particular command, please
open an issue on the rtx repo. You can also setup a shell function to reshim
automatically (it's really fast so you don't need to worry about overhead):
npm() {
command npm "$@"
rtx reshim
}
Usage: reshim
Examples:
$ rtx reshim
$ ~/.local/share/rtx/shims/node -v
v20.0.0
Updates rtx itself
Uses whatever package manager was used to install rtx or just downloads
a binary from GitHub Releases if rtx was installed manually.
Supports: standalone, brew, deb, rpm
Usage: self-update
Show a current setting
This is the contents of a single entry in ~/.config/rtx/config.toml
Note that aliases are also stored in this file
but managed separately with `rtx aliases get`
Usage: settings get <KEY>
Arguments:
<KEY>
The setting to show
Examples:
$ rtx settings get legacy_version_file
true
Show current settings
This is the contents of ~/.config/rtx/config.toml
Note that aliases are also stored in this file
but managed separately with `rtx aliases`
Usage: settings ls
Examples:
$ rtx settings
legacy_version_file = false
Add/update a setting
This modifies the contents of ~/.config/rtx/config.toml
Usage: settings set <KEY> <VALUE>
Arguments:
<KEY>
The setting to set
<VALUE>
The value to set
Examples:
$ rtx settings set legacy_version_file true
Clears a setting
This modifies the contents of ~/.config/rtx/config.toml
Usage: settings unset <KEY>
Arguments:
<KEY>
The setting to remove
Examples:
$ rtx settings unset legacy_version_file
Sets a tool version for the current shell session
Only works in a session where rtx is already activated.
Usage: shell [OPTIONS] [TOOL@VERSION]...
Arguments:
[TOOL@VERSION]...
Tool(s) to use
Options:
-u, --unset
Removes a previously set version
Examples:
$ rtx shell node@20
$ node -v
v20.0.0
Symlinks all tool versions from an external tool into rtx
For example, use this to import all Homebrew node installs into rtx
Usage: sync node <--brew|--nvm|--nodenv>
Options:
--brew
Get tool versions from Homebrew
--nvm
Get tool versions from nvm
--nodenv
Get tool versions from nodenv
Examples:
$ brew install node@18 node@20
$ rtx sync node --brew
$ rtx use -g node@18 - uses Homebrew-provided node
Symlinks all tool versions from an external tool into rtx
For example, use this to import all pyenv installs into rtx
Usage: sync python --pyenv
Options:
--pyenv
Get tool versions from pyenv
Examples:
$ pyenv install 3.11.0
$ rtx sync python --pyenv
$ rtx use -g [email protected] - uses pyenv-provided python
Marks a config file as trusted
This means rtx will parse the file with potentially dangerous
features enabled.
This includes:
- environment variables
- templates
- `path:` plugin versions
Usage: trust [OPTIONS] [CONFIG_FILE]
Arguments:
[CONFIG_FILE]
The config file to trust
Options:
--untrust
No longer trust this config
Examples:
# trusts ~/some_dir/.rtx.toml
$ rtx trust ~/some_dir/.rtx.toml
# trusts .rtx.toml in the current or parent directory
$ rtx trust
Removes runtime versions
Usage: uninstall <TOOL@VERSION>...
Arguments:
<TOOL@VERSION>...
Tool(s) to remove
Examples:
$ rtx uninstall [email protected] # will uninstall specific version
$ rtx uninstall node # will uninstall current node version
[experimental] Upgrades outdated tool versions
Usage: upgrade [TOOL@VERSION]...
Arguments:
[TOOL@VERSION]...
Tool(s) to upgrade
e.g.: node@20 [email protected]
If not specified, all current tools will be upgraded
Change the active version of a tool locally or globally.
This will install the tool if it is not already installed.
By default, this will use an `.rtx.toml` file in the current directory.
Use the --global flag to use the global config file instead.
This replaces asdf's `local` and `global` commands, however those are still available in rtx.
Usage: use [OPTIONS] [TOOL@VERSION]...
Arguments:
[TOOL@VERSION]...
Tool(s) to add to config file
e.g.: node@20
If no version is specified, it will default to @latest
Options:
--pin
Save exact version to config file
e.g.: `rtx use --pin node@20` will save `node 20.0.0` to ~/.tool-versions
--fuzzy
Save fuzzy version to config file
e.g.: `rtx use --fuzzy node@20` will save `node 20` to ~/.tool-versions
this is the default behavior unless RTX_ASDF_COMPAT=1
--remove <TOOL>
Remove the tool(s) from config file
-g, --global
Use the global config file (~/.config/rtx/config.toml) instead of the local one
-p, --path <PATH>
Specify a path to a config file or directory
Examples:
# set the current version of node to 20.x in .rtx.toml of current directory
# will write the fuzzy version (e.g.: 20)
$ rtx use node@20
# set the current version of node to 20.x in ~/.config/rtx/config.toml
# will write the precise version (e.g.: 20.0.0)
$ rtx use -g --pin node@20
Show rtx version
Usage: version
Display the installation path for a runtime
Must be installed.
Usage: where <TOOL@VERSION>
Arguments:
<TOOL@VERSION>
Tool(s) to look up
e.g.: ruby@3
if "@<PREFIX>" is specified, it will show the latest installed version
that matches the prefix
otherwise, it will show the current, active installed version
Examples:
# Show the latest installed version of node
# If it is is not installed, errors
$ rtx where node@20
/home/jdx/.local/share/rtx/installs/node/20.0.0
# Show the current, active install directory of node
# Errors if node is not referenced in any .tool-version file
$ rtx where node
/home/jdx/.local/share/rtx/installs/node/20.0.0
Shows the path that a bin name points to
Usage: which [OPTIONS] <BIN_NAME>
Arguments:
<BIN_NAME>
The bin name to look up
Options:
--plugin
Show the plugin name instead of the path
--version
Show the version instead of the path
-t, --tool <TOOL@VERSION>
Use a specific tool@version
e.g.: `rtx which npm --tool=node@20`
Examples:
$ rtx which node
/home/username/.local/share/rtx/installs/node/20.0.0/bin/node
$ rtx which node --plugin
node
$ rtx which node --version
20.0.0