Installation
Installing Devlish gives you a command-line application for running .dvl programs. The compiler and runtime are part of the application, so running a workflow does not require a separate compiler command or an AI account.
The current repository installs from source using its Makefile. You need Git, a Rust toolchain with Cargo, and make for the commands below. These steps use a macOS or Linux terminal. The browser playground is also available if you want to try the language before setting up build tools.
Download and install
Section titled “Download and install”Clone the repository so the installer has the source and build configuration it needs. make install builds the release binary using the lockfile and installs it as devlish:
git clone https://github.com/adubinsky/devlish.gitcd devlishmake installThe default destination is ~/.local/bin/devlish. Make sure ~/.local/bin is on your PATH, the list of directories your terminal searches for commands. You can choose another prefix with make install PREFIX=/usr/local if you have permission to install there.
The repository’s install.sh delegates to this same build-and-install process. It is not a standalone download of a prebuilt binary.
Check the application
Section titled “Check the application”Version output confirms the executable is available. Help describes the ordinary application modes and their optional settings:
devlish --versiondevlish --helpIf your terminal cannot find devlish, check that the installation directory is on PATH, then open a new terminal or reload your shell configuration.
Build without installing
Section titled “Build without installing”Contributors who want a local executable can build it without copying it to an installation directory:
make buildbin/devlish --helpThe native binary is built under crates/devlish_core/target/release/. The repository’s bin/devlish wrapper locates the development binary.
Run your first program
Section titled “Run your first program”Continue to Quick Start for a small program that reads an input and produces a result. Your First Tool then shows how an AI application can call a program through MCP.