Loading
sir.sh
Local-first workflow automation for AI-assisted development
A local-first workflow runner built on Laravel Zero that turns high-level instructions into executable YAML workflows, with layered `.sir` configuration, reusable Git-installed packs, built-in tasks, and an MCP-style interface for AI assistants.
Tag
Developer Tools · Workflow Runner
Duration
Ongoing
Role
Product framing · Frontend · CLI UX
sir.sh sells the right mental model fast: define workflows once, keep them local-first, layer configuration by directory, install packs from Git, and let AI assistants discover and execute automation safely.
Execution model
YAML workflows
Runtime
Laravel Zero
Config
Layered .sir
◇ The problem
Where they were stuck.
Developers needed something more structured than ad hoc shell scripts but lighter than a full orchestration platform. sir.sh had to make complex workflows portable, inspectable, and AI-friendly without giving up local control, dry-run safety, or repo-level customization.
◇ Our approach
How we shipped it.
- 01
Framed the product around a practical developer workflow: define once in YAML, resolve local and global `.sir` layers automatically, then run, preview, or share the same automation surface across projects.
- 02
Made the architecture legible through the marketing site and source story together, highlighting the core pieces that matter in practice: layered configuration, workflow engine, pack system, built-in tasks, and context-aware execution.
- 03
Positioned the AI agent interface as a controlled extension of the workflow runner rather than an opaque autopilot, so assistants can discover tools, inspect plans, and execute workflows with explicit context and safer boundaries.
- 04
Kept the product grounded in real CLI ergonomics: clean output, reusable methods, Git-based pack distribution, spec coverage, and a local-first model that fits how developers actually work.
◇ Stack
What it's made of.
PHPLaravel ZeroYAML workflowsGit pack systemCLI task runnerMCP-style agent tools
Next case study