OpenSpec
OpenSpec is an open-source spec-driven development toolkit designed specifically for AI coding assistants. It extends the spec-before-code pattern with machine-readable conventions that allow AI agents to read a spec, implement against it, and mark sections complete as work progresses. Unlike general SDD tooling, OpenSpec's format is optimised as a structured work order for AI agents rather than as human-readable design documentation.
spec-kit (already on this radar) formalises the practice of writing plain-English specifications before implementation. OpenSpec addresses a different concern: how to structure those specifications so that AI coding agents can consume them reliably. The format includes machine-readable sections defining what to implement, what constraints apply, and how to verify completion — turning the spec into an executable instruction set for the agent rather than a reference document for a human reviewer.
Placed in Assess at center position because AI coding agent workflows are maturing and the lack of structured input formats is a real source of drift and incomplete implementations. OpenSpec is an early but concrete response to that gap. The center position reflects that the tool is still gaining adoption and the convention space for AI-native spec formats has not yet converged; teams experimenting with agent-driven development should evaluate it alongside spec-kit rather than as a replacement.
The distinction from spec-kit: spec-kit targets human design intent and review; OpenSpec targets agent instruction and verification.
