Verifiable Architecture Visualization: Meet Archify
As autonomous AI coding assistants (such as Claude Code, Cursor, and Codex CLI) become central to system design, engineering teams increasingly use them to map complex architectures. However, typical AI-drawn diagrams suffer from inconsistent geometry, untyped syntax errors, and an inability to track structural changes across Git revisions.
Archify is an open-source diagramming and validation engine developed by tt-a1i to bring rigor to AI-generated system maps. Rather than generating loose markdown charts, Archify requires AI agents to produce a typed JSON Intermediate Representation (IR) that compiles deterministically into interactive, self-contained HTML and SVG artifacts.
What is Archify?
Archify operates as a verification engine and rendering compiler. When you ask an AI agent to map a codebase or design a cloud architecture, the agent outputs a structured JSON schema. Archify validates node clearances, boundary crossings, and layout hierarchies before generating a complete, standalone visual artifact.
Key Core Features
1. Five Specialized Diagram Types
Archify supports five core technical visualization models:
- Architecture: Component services, databases, external dependencies, and trust boundaries.
- Workflow: Multi-lane CI/CD pipelines, approvals, runbooks, and exception handlers.
- Sequence: API call chains, authentication flows, cache fallbacks, and async event traces.
- Data Flow: Data pipelines, ETL transforms, storage tiers, and PII boundaries.
- Lifecycle: Finite state machines, retries, timeout loops, and terminal states.
2. Architecture Delta Review
During pull request reviews or system refactors, Archify supports side-by-side snapshot diffing. Developers can compare Before, Delta, and After states to inspect exact added, removed, moved, or rerouted components with a deterministic verification receipt.
3. Interactive Standalone HTML Viewer
Archify outputs self-contained HTML files with advanced interactive capabilities:
-
Route Probing (
R): Inspect the shortest authored path between services. - Upstream/Downstream Reach: Highlight exact dependencies without claiming unverified runtime impact.
-
Presentation Stage (
F): Present guided architecture walkthroughs chapter by chapter. - Export Options: Download high-resolution SVG, PNG, WebM animations, or canonical 1200×630 social share cards.
4. Universal Agent Skill Integration
Archify installs seamlessly across modern AI coding environments:
# Global installation for AI coding agents
npx skills add tt-a1i/archify -g
# Direct Cursor install
npx -y skills add tt-a1i/archify --skill archify --agent cursor --global --copy --yes
Conclusion
By combining typed JSON validation with deterministic vector compilation, Archify transforms natural language system descriptions into trustworthy, publication-ready architecture diagrams. It is an indispensable tool for senior engineers and system architects pair-programming with AI agents.
Want to turn your architecture descriptions into verifiable system maps? Check out the Archify GitHub Repository.
Top comments (1)
The typed JSON IR and deterministic HTML/SVG compilation give architecture diagrams a testable boundary that loose Markdown charts usually lack. The Before/Delta/After review and route probing are especially useful, though I'd want every meaningful node and edge tied to a source anchor-code, config, or an explicit human assertion-so a valid diagram cannot quietly become a confidently rendered guess. Geometry checks and typed syntax prove the artifact is internally sound; provenance is what makes it dependable during a real refactor.