Skip to main content

Reference

📄️Commands Reference

Sooner or later on any engagement you will want to know exactly what Wire ran on your behalf, or you will want to run one step yourself rather than ask for it, and for either of those you need the names. This page is the catalogue of them. Every Wire command follows the pattern /wire:- , and every artifact has three lifecycle commands, generate, validate and review, except where noted in the tables below, so that once you know an artifact's name you can usually work out its commands without looking them up.

📄️Skills Reference

Much of the day-to-day work on an engagement is not a Wire command at all: you are writing a dbt model, fixing a LookML view or working out why a job failed overnight, and what you want is for the conventions Wire encodes to apply to that work without your having to look them up. That is what skills are for. Wire skills are Markdown files that Claude loads as instruction sets, and unlike commands, which you invoke explicitly, most skills auto-activate, which is to say they "fire" when the work you are doing matches their trigger conditions, without you needing to reference them.

📄️MCP Servers

Much of what matters on an engagement happens outside the repository: decisions are taken on client calls, progress is tracked in Jira or Linear and documents are read and commented on in Confluence or Notion, and if Wire could see none of that its record would be a thinner thing. Wire integrates with five MCP (Model Context Protocol) servers to close that gap. All are optional, and Wire works without any of them, but when they are present they add meeting context to reviews, sync artifact status to issue trackers, replicate artifacts to document stores and provide library documentation lookups during development.