Quick Start
Meriadoc is a task runner for both humans and AI agents. You define tasks, jobs, and shell environments in a YAML file that lives in your project, then run them from anywhere on your machine.
Installation
Homebrew (macOS)
brew tap meriadoc-dev/meriadoc
brew install meriadoc
One-line install script (macOS and Linux)
curl -fsSL https://raw.githubusercontent.com/meriadoc-dev/meriadoc/main/install.sh | sh
Cargo
cargo install meriadoc
Prebuilt binaries for Linux (gnu, musl), macOS (arm64, x86_64), and Windows are available on GitHub Releases.
Create a spec file
Create meriadoc.yaml in your project root:
version: v1
tasks:
build:
description: "Build the project"
cmds:
- cargo build --release
test:
description: "Run tests"
cmds:
- cargo test
deploy:
description: "Deploy to an environment"
agent:
risk_level: high
confirmation: "This will deploy to production. Continue?"
cmds:
- ./deploy.sh
env:
ENVIRONMENT:
type: choice
options: [dev, staging, prod]
default: dev
jobs:
ci:
description: "Full CI pipeline"
tasks:
- build
- test
shells:
dev:
description: "Development environment"
env:
RUST_LOG:
type: string
default: debug
Register your project
meriadoc config add /path/to/your/project
Meriadoc discovers spec files by walking registered directories. You can add as many project roots as you want.
Run your first task
meriadoc ls tasks # See what's available
meriadoc run task build # Run the build task
meriadoc run job ci # Run the ci job
meriadoc run shell dev # Start the dev shell
Shortcuts
Every run task, run job, and run shell has a short form:
meriadoc task build # same as: meriadoc run task build
meriadoc t build # even shorter
meriadoc job ci # same as: meriadoc run job ci
meriadoc j ci
meriadoc shell dev # same as: meriadoc run shell dev
meriadoc s dev
Web UI
Start the web server for a browser-based interface:
meriadoc server
Opens at http://localhost:8420. Browse all projects, inspect task details, run with typed inputs, and watch real-time output.
What’s next
- Tasks — learn about env vars, preconditions, and failure handlers
- Jobs — compose tasks into multi-step workflows
- CLI Reference — all commands and flags
- Agents / MCP — integrate with Claude and other AI agents