Archify
archify is an agent-driven diagramming tool that turns a plain-English description of a system, or a codebase itself, into an interactive technical diagram. It supports five diagram types: architecture, workflow, sequence, data-flow, and lifecycle. Output is a single self-contained HTML file with SVG/PNG export, motion, and light/dark themes, so a diagram can be shared or embedded without any server or viewer app.
The tool is designed to be driven by a coding agent rather than a human clicking a diagramming UI. A supporting agent (Claude Code, Cursor, Codex CLI, and others are supported) generates a typed JSON intermediate representation of the system, bundled checkers validate that representation, and only a version that passes validation is rendered or allowed to replace an existing diagram. This generate-validate-deliver loop is meant to prevent an agent from silently producing a diagram that looks plausible but does not match the underlying system. archify also has an "Architecture Delta" mode that compares two snapshots of a system and highlights what changed, with a machine-readable record of the comparison.
archify explicitly does not attempt general-purpose diagram layout, automatic parsing of existing Mermaid diagrams, hosted sharing, or a WYSIWYG editor -- it is scoped to agent-generated, validated diagrams only. As of research, the project has around 50,000 GitHub stars and is under active development, with 54 open issues tracked.
archify sits in the same problem space as Agent-Driven Draw.io Diagramming and diagram-design, both of which also use an agent to produce diagrams for a codebase. The difference is output format: archify produces a self-contained interactive HTML artifact with built-in validation against the source, rather than editable native diagram files (.drawio) or static HTML+SVG templates.
archify is placed in Assess. It has not yet been used first-person in any production pipeline; its ring reflects genuine promise and a large, active open-source community rather than confirmed value in practice. The validation loop (generate, check, only deliver a passing artifact) is a notable design choice worth exploring further, since it addresses a common failure mode of agent-generated diagrams drifting from the real system they describe.
