Installation & Distribution
CURD provides native artifacts, language-bound wrappers, and OS-level package manifests to fit any development environment.
We distribute our binaries via two primary paths: Native OS Packages (for pure command-line use) and Wrapper Launcher Ecosystems (for using CURD within Node or Python automation scripts).
1. Native OS Package Managers
These are the recommended installation methods if you simply want the curd binary available on your $PATH.
Homebrew (macOS & Linux)
We maintain a Homebrew tap for Apple Silicon, Intel Macs, and Linux.
brew tap aerospeedsta/curd
brew install curdWinget (Windows)
For Windows 10/11 users, you can install the standalone executable natively:
winget install curdScoop (Windows)
scoop bucket add curd https://github.com/aerospeedsta/curd-scoop.git
scoop install curdArch Linux (AUR)
If you are on Arch Linux or an Arch derivative (Manjaro, EndeavourOS):
paru -S curd-bin
# or
yay -S curd-binDebian/Ubuntu & RHEL/Fedora
We provide pre-built .deb and .rpm packages attached to the GitHub Releases page.
# Debian/Ubuntu
wget https://github.com/aerospeedsta/CURD/releases/latest/download/curd_amd64.deb
sudo dpkg -i curd_amd64.deb
# RHEL/Fedora
wget https://github.com/aerospeedsta/CURD/releases/latest/download/curd_x86_64.rpm
sudo rpm -i curd_x86_64.rpm2. Wrapper & Launcher Ecosystems
If you are using tools like uvx, npx, or bunx to drive automation, you don’t need a system-level install. You can fetch and execute CURD on the fly using our native bindings.
(These wrappers automatically detect your OS/architecture and pull down the correct pre-compiled Rust binary.)
Python-First Launchers (PyPI)
If your agents or scripts are Python-based, use curd-python:
# Run instantly with uv (Recommended)
uvx --from curd-python curd --version
# Or via pipx
pipx install curd-pythonNode-First Launchers (npm)
If your environment is Node/TS based, use curd-node:
# Run instantly with bun
bunx --bun curd-node --version
# Or via npx
npx curd-node --version3. Sandboxed Ecosystems
Nix / NixOS
CURD provides a flake.nix making it directly accessible to Nix environments.
nix run github:aerospeedsta/CURDMise (Formerly RTX)
Add this to your mise.toml:
[tools]
"cargo:curd" = "latest"Pixi
Add CURD as a global pixi task or tool:
pixi global install curd-python4. Containerized Execution (Docker)
If you prefer to run CURD in a completely isolated container, or if you want to serve the CURD MCP server without installing anything on your host machine, you can use our official Docker image.
# Run the CURD MCP server in the current directory
docker run -it --rm -v $(pwd):/workspace aerospeedsta/curd:latestThe image defaults to curd mcp ., serving the Model Context Protocol over stdio. It is a minimal Alpine-based image (~15MB) containing statically linked binaries for both x86_64 and arm64.
5. Manual Pre-compiled Binaries
You can always download standalone, statically linked binaries directly from the GitHub Releases page. We provide:
x86_64-unknown-linux-musl(Static Linux, great for minimal Docker containers like Alpine)aarch64-unknown-linux-musl(Static Linux ARM64)x86_64-apple-darwin(macOS Intel)aarch64-apple-darwin(macOS Silicon)x86_64-pc-windows-msvc(Windows x64)