Concepts
Specwright is built on a few core ideas that distinguish it from traditional documentation tools. Understanding these concepts will help you get the most out of the platform.
Core Thesis
Documentation fails because it's a write-once artifact. Specwright treats docs as living programs — all markdown is input, AI agents are the runtime, code is the ground truth, and the repo is the execution environment.
Specs define intent — what should be built. Code reveals reality — what actually shipped. Other docs (ADRs, guides, READMEs) provide context — why decisions were made. The platform closes the loop across all three, continuously evaluating whether intent matches reality and keeping context accurate.
Key Concepts
OpenSpec Framework
The structured artifact model that makes specs machine-readable. Proposals, sections, acceptance criteria, and tasks — all in markdown, all tracked.
Living Specs
The feedback loop that keeps documentation alive. PMs write specs, agents create tickets, engineers write code, agents verify implementation, docs auto-update.
Delta Tracking
How Specwright tracks changes through status transitions and hidden comments rather than rewriting specs.
Agent Mesh
The agent architecture — from single-repo agents to cross-repo coordination to the org-wide knowledge brain.
Spec Coverage
The metric model that measures what percentage of spec acceptance criteria are verified against actual code.