Xacrodoc
Compile xacro files to plain URDF or MJCF from Python or the command line (no ROS required).
Install / Use
/learn @adamheins/XacrodocREADME
xacrodoc
xacrodoc is a tool for compiling xacro files to plain URDF files or Mujoco MJCF files from within Python code or via the command line. It is fully functional whether ROS is installed on the system or not.
Why?
Compared to the regular xacro package, xacrodoc allows you to:
- Conveniently convert xacro files to URDF or Mujoco MJCF files without a ROS installation (this includes converting plain URDF to MJCF).
- Programmatically compose multiple xacro files and apply substitution arguments to build a flexible URDF model directly in your code. This allows you to avoid the clutter of redundant compiled raw URDFs and only keep the xacro source files.
- Seamlessly obtain URDF strings and (temporary) URDF file paths as needed. For example, many libraries (such as Pinocchio) accept a URDF string to build a model, but others (like PyBullet) only load URDFs directly from file paths.
- Easily copy assets like meshes to a single directory and have fine-grained control over file protocols and absolute vs. relative paths, if desired.
Documentation
Usage information and examples are provided in this README. The package's full API reference can be found here.
Installation
xacrodoc requires at least Python 3.8. Note that ROS does not need to be installed on the system, but xacrodoc will also use its infrastructure to look for packages if it is available.
Command Line Tool
It is recommended to install the command line tool into an isolated environment using uv:
uv tool install xacrodoc
# for conversion to MJCF files, use:
uv tool install "xacrodoc[mujoco]"
or pipx:
pipx install xacrodoc
# or
pipx install "xacrodoc[mujoco]"
Python Library
The library can be installed from pip:
pip install xacrodoc
# or to also include conversion to MJCF files:
pip install "xacrodoc[mujoco]"
or from source:
git clone --recurse-submodules https://github.com/adamheins/xacrodoc
cd xacrodoc
pip install .
Command Line Usage
In addition to the Python API described below, this package also includes a
xacrodoc command line tool. It is similar to xacro but provides additional
options; notably, directories in which to search for packages can be provided
manually. Examples:
# compile and print to stdout
xacrodoc input.urdf.xacro
# compile and output to provided output file
xacrodoc input.urdf.xacro -o output.urdf
# provide directories in which to look for packages referenced in
# input.urdf.xacro (the -d (--pkg-dir) flag is needed before each one to
# disambiguate from substitution arguments); a common use-case would be
# providing the path to catkin workspaces
xacrodoc input.urdf.xacro -d ~/my_pkg_dir -d ~/my_other_pkg_dir
# alternatively, package name/path mappings can be directly supplied using the
# -p (--pkg-path) flag; := is used to separate the name and path:
xacrodoc input.urdf.xacro -p my_pkg:=~/path/to/my_pkg -p other_pkg:=/home/foo/packages/other_pkg
# substitution arguments also use := notation, like xacro
xacrodoc input.urdf.xacro mass:=1
# you can also remove file:// protocols prefixed to asset paths with
# -s (--strip-protocols) or convert paths from absolute to relative using
# -r (--relative-paths)
xacrodoc input.urdf.xacro -s -r output.urdf
# alternatively, convert to MJCF (requires Mujoco)
# the -c (--copy-assets-to) option with a directory name is required when
# converting to MJCF; it copies all assets to that directory and updates their
# file paths (-c is optional when converting to URDF)
# relative file paths are always used when converting to MJCF
xacrodoc input.urdf.xacro --mjcf -c assets -o output.xml
Python Library Usage
Basic
A basic use-case of compiling a URDF from a xacro file:
import os
from xacrodoc import XacroDoc
doc = XacroDoc.from_file("robot.urdf.xacro")
# or relative to a ROS package
# e.g., for a file located at some_ros_package/urdf/robot.urdf.xacro:
doc = XacroDoc.from_package_file("some_ros_package", "urdf/robot.urdf.xacro")
# convert to a string of URDF
urdf_str = doc.to_urdf_string()
# or write to a file
doc.to_urdf_file("robot.urdf")
# or just work with a temp file
# this is useful for working with libraries that expect a URDF *file* (rather
# than a string)
with doc.temp_urdf_file_path() as path:
# do stuff with URDF file located at `path`...
# file is cleaned up once context manager is exited
# you can also manage the temp file yourself if you don't want to clean it up
# right away
path = doc.to_temp_urdf_file()
# ...do stuff with path...
# manually delete the temp file
os.remove(path)
Finding ROS packages
xacro files often make use of $(find <pkg>) directives to resolve paths
relative to a given ROS package. If ROS is installed on the system, xacrodoc
automatically looks for ROS packages using the usual ROS infrastructure. If
not, or if you are working with packages outside of a ROS workspace, you'll
need to tell xacrodoc where to find packages. There are a few ways to do this:
import xacrodoc as xd
# `from_file` automatically resolves packages by looking in each parent
# directory of the given path to check for required ROS packages (as marked by
# a package.xml file)
doc = xd.XacroDoc.from_file("robot.urdf.xacro")
# if you want to disable this, pass `walk_up=False`:
doc = xd.XacroDoc.from_file("robot.urdf.xacro", walk_up=False)
# we can also tell xacrodoc to walk up a directory tree manually
xd.packages.walk_up_from("some/other/path")
# or we can give paths to directories to search for packages
# packages can be located multiple levels deep from the specified directories,
# just like in a ROS workspace - the same package search logic is used (since
# we actually use rospkg under the hood)
xd.packages.look_in(["somewhere/I/keep/packages", "another/directory/with/packages"])
# you can bypass ROS conventions entirely by directly providing a dict of
# package names and paths; no package.xml is required to identify packages in
# this case
xd.packages.update_package_cache({"my_package": "path/to/my_package"})
Multiple URDFs
We can also build a URDF programmatically from multiple xacro files:
import xacrodoc as xd
# setup where to look for packages, if needed; for example:
xd.packages.look_in(["somewhere/I/keep/packages"])
# specify files to compose (using xacro include directives)
includes = ["robot_base.urdf.xacro", "robot_arm.urdf.xacro", "tool.urdf.xacro"]
doc = xd.XacroDoc.from_includes(includes)
# includes can also use $(find ...) directives:
includes = [
"$(find my_ros_package)/urdf/robot_base.urdf.xacro",
"$(find another_ros_package)/urdf/robot_arm.urdf.xacro",
"tool.urdf.xacro"
]
doc = xd.XacroDoc.from_includes(includes)
Substitution arguments
We can also pass in substitution arguments to xacro files. For example, suppose our
file robot.urdf.xacro contains the directive <xacro:arg name="mass" default="1"/>.
On the command line, we could write
xacro robot_base.urdf.xacro -o robot_base.urdf mass:=2
to set the mass value. Programmatically, we do
from xacrodoc import XacroDoc
doc = XacroDoc.from_file("robot.urdf.xacro", subargs={"mass": "2"})
Resolving file names with respect to packages
One feature of URDF (not just xacro files) is that file names (e.g., for meshes) can be specified relative to a package by using
package://<pkg>/relative/path/to/mesh
syntax, which depends on ROS and is not supported by other non-ROS tools. xacrodoc automatically expands these paths out to full absolute paths, e.g.,
file:///abs/path/to/mesh
but this can be disabled by passing resolve_packages=False to the XacroDoc
constructor methods.
Additional file name manipulation
As just described, URDFs typically prefix asset file names with a protocol,
either file:// or package://, and the resolution of the latter to the
former can be controlled using the resolve_packages parameter.
For some applications, one may wish to eliminate the protocol prefixes
entirely. This can be done by passing use_protocols=False to the
XacroDoc.to_urdf_string or XacroDoc.to_urdf_file methods, which will strip
away any file:// protocols (it doesn't make sense to strip an unresolved
package:// prefix, an error is raised if one is found when
use_protocols=False). Conversely, if use_protocols=True (the default), then
a file:// protocol prefix is added to any file names without a prefix
and package:// prefixes are left unchanged. Note that protocols are always
stripped when converted to MJCF.
In addition, one can control whether resolved file paths are absolute or
relative. By default, URDF works with absolute paths, but many applications
support relative paths, which are also more portable. When converting to a file
using the XacroDoc.to_urdf_file method, passing relative_paths=True will
make all file names relative to the output file location. When converting to a
string using XacroDoc.to_urdf_string, you can pass a path to
paths_relative_to to make all file names relative to that path in the
resulting URDF string.
When loading a document that has relative file names, they are resolved
relative to the XacroDoc.rootdir attribute if it is not None. When the
document is loaded from a file using XacroDoc.from_file or
XacroDoc.from_package_file, rootdir is automatically set to be that file's
parent directory. When the document is loaded using XacroDoc.from_string or
XacroDoc.from_includes, rootdir can be specified as an argument.
Conversion to MJCF format
Mujoco has basic support for URDFs, but natively uses its own MJCF XML format. If you want to use Mujoco, you probably
Related Skills
node-connect
333.3kDiagnose OpenClaw node connection and pairing failures for Android, iOS, and macOS companion apps
claude-opus-4-5-migration
82.0kMigrate prompts and code from Claude Sonnet 4.0, Sonnet 4.5, or Opus 4.1 to Opus 4.5
frontend-design
82.0kCreate distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.
model-usage
333.3kUse CodexBar CLI local cost usage to summarize per-model usage for Codex or Claude, including the current (most recent) model or a full model breakdown. Trigger when asked for model-level usage/cost data from codexbar, or when you need a scriptable per-model summary from codexbar cost JSON.
