From feature request to spec, automatically

Label an issue and Overcut goes to work: it identifies the relevant repositories, clones them, and runs a deep codebase analysis to map patterns and constraints. A Senior Developer and System Architect then draft the requirements through an iterative review loop, and a Product Manager posts the finished document straight to the ticket.

The edge cases, covered

Requirements only matter if they hold up. Every document is grounded in real code, reviewed for completeness, and verified before it ever reaches your ticket.

Grounded in real code

A deep codebase analysis maps affected components, existing patterns, and technical constraints with file-level references — so requirements reflect reality, not guesswork.

Iterative agent review

A Senior Developer drafts the spec and a System Architect reviews it for completeness and feasibility, looping up to two revisions until the document is sound.

Open questions surfaced early

Ambiguities and gaps are flagged as explicit open questions, so the team can resolve scope before any design or implementation work begins.

Smart labeling on completion

Tickets are labeled `requirements-complete` when the spec is clean, or `requirements-needs-info` when open questions remain — keeping your board honest.

Verified before posting

Each stage confirms its output exists and contains the required sections before proceeding, so an incomplete or empty document never lands on your ticket.

Built on your repos and tracker

Overcut reads your issues, clones your repositories, and writes requirements back as comments and labels in the tools your team already lives in.

Running from day one

Runs automatically the moment an issue gets the `needs-requirements` label, or on demand with the `/requirements` command on any ticket. The only setup is configuring your Senior Developer, Product Manager, and System Architect agents.