Linkar documentation

Linkar is a lightweight runtime for reusable computational templates. It helps teams publish scripts, CLI tools, notebooks, and analysis steps as Git-backed packs with traceable local runs for humans, scripts, and agents.

The core idea is deliberately small: packs distribute reusable templates and bindings, projects record local work, and every run leaves readable artifacts on disk. Linkar does not try to become a workflow engine. It provides the contract around shared analysis building blocks.

Installation

Install Linkar into the environment where you want to manage packs and project runs.

pipx install git+https://github.com/chaochungkuo/linkar.git

After installation, add a pack and initialize a project:

linkar config pack add github:IZKF-Genomics/izkf_pack --id izkf_pack
linkar project init --name study
cd study

Public interfaces

Linkar exposes the same runtime semantics through three stable user-facing interfaces.

CLI Run, render, collect, inspect, clean, and manage projects from the terminal. Local API Resolve templates and read run metadata through local HTTP endpoints. MCP Expose Linkar operations to agent clients without shell scraping.

Documentation

Start with the guides if you are learning by doing. Use the conceptual pages when you need the runtime model, project ledger, cleanup contract, or automation boundary.

Installation and first run Create a project, run a template, and inspect outputs. Template contract Understand params, outputs, cleanup rules, tools, and run modes. Examples Choose a bundled example pack based on the pattern you want to inspect.

Project model

A Linkar project is a local directory with a project ledger, stable visible run paths, and optional immutable run history. This keeps analysis handoffs readable while still retaining enough metadata for provenance.

linkar render demultiplex --outdir ./demultiplex
bash ./demultiplex/run.sh
linkar collect ./demultiplex
linkar clean .