Local-first workflow automation for AI-assisted development
sir.sh is a local-first workflow runner built on Laravel Zero. It turns instructions into YAML workflows, with layered `.sir` config, Git-installed packs, built-in tasks and an MCP-style interface for AI assistants.
The sir.sh site showing YAML workflows, config layered by directory, packs installed from Git, and AI assistants running workflows.
Where they were stuck.
Developers wanted more structure than ad hoc shell scripts, without a full orchestration platform. sir.sh makes workflows portable and inspectable, works with AI assistants, and keeps local control, dry runs and per-repo config.
How we shipped it.
- 01
Wrote the site around the basic flow. Define a workflow in YAML, let sir.sh merge local and global `.sir` layers, then run, preview or share it.
- 02
Explained the main parts on the marketing site: layered configuration, the workflow engine, packs, built-in tasks and context-aware execution.
- 03
Described the AI interface as part of the runner. Assistants can discover tools, inspect plans and run workflows with explicit context and limits.
- 04
Showed the CLI as it works day to day, with clean output, reusable methods, packs shared through Git, spec coverage and local execution.