Hal is a terminal-first CLI that runs PRD-driven development loops with AI coding agents, giving each story a structured path through planning, validation, implementation, review, reporting, and archive state.
Hal is for developers who want AI agents to execute product work without turning a feature into one long, unstructured session.
How Hal works
A practical Hal workflow starts with requirements. The developer creates or imports a PRD, converts it into structured runtime state, validates the stories, and then runs implementation iterations.
Each iteration can pick the next story, load project standards, ask an engine such as Codex, Claude Code, or Pi to implement the change, commit work, update progress, and leave a report for review.
When Hal fits
Hal fits when a solo developer or technical founder wants agent speed with a more inspectable process: small stories, fresh context, standards, reports, CI hooks, and archive behavior instead of a single giant prompt.
FAQ
What is Hal?
Hal is a CLI for autonomous PRD-driven coding loops. It helps an AI coding agent work through scoped stories with fresh context, project standards, commits, reports, and reviewable state.
Related pages
Next step
Run autonomous PRD-driven coding loops with AI agents.
View Hal on GitHub