cd /news/ai-agents/show-hn-arkviz-ai-generated-software… · home › topics › ai-agents › article
[ARTICLE · art-173451] src=arkviz.io ↗ pub= topic=ai-agents verified=true sentiment=↑ positive

Show HN: ArkViz – AI generated software architecture diagrams

Bryant, founder of ArkViz, launched ArkViz, a product that uses AI to automatically generate software architecture diagrams from a codebase, built on the open-source LikeC4 architecture-as-code format. ArkViz pitches its structured model diagrams as deeper context for AI agents than text files like CLAUDE.md, AGENTS.md and ARCHITECTURE.md, and adds hosting, access control, scheduling automation, error handling and retries, logs and GitHub connectors that Bryant says users would otherwise have to build themselves with a harness like Claude Code.

key takeaway

ArkViz generates structured LikeC4 model diagrams from a codebase so AI agents get a single-source-of-truth architecture context rather than pixel-based canvas drawings or loosely structured Mermaid files.

read3 min views4 publishedNov 20, 2026

Hi HN, I’m Bryant, founder of ArkViz (https://arkviz.io). ArkViz is a product to automatically generate architecture diagrams from your codebase using AI.

The vision behind ArkViz is to supercharge both AI agents and humans by providing AI agents deeper relational context using structured model diagram formats, and by providing humans with a rich visual representation you can interact with, with features like zooming, panning, search, and show composing files.

In this new age of AI, you can generate a structured model diagram from your codebase in seconds, quicker and better than a human. Not only can the generated diagram be a visual documentation tool for humans, AI agents can also benefit from the context of a pre-generated architecture diagram that lives alongside a project’s codebase. This concept has already been proven by context files like CLAUDE.md, AGENTS.md, ARCHITECTURE.md living in the root of your repository.

The differentiator between ARCHITECTURE.md and ArkViz is that ArkViz generates a structured model diagram of the system, while ARCHITECTURE.md is broader in scope and purely text. ArkViz aims to be a supplement to ARCHITECTURE.md to provide deeper understanding like how components are connected, which components live inside other components, which files compose a component.

—

Why ArkViz?

Software engineering teams often create diagrams to aid in the documentation of their software systems. They are incredibly useful for onboarding new engineers and understanding projects. However, they take time to create and maintain, so are often non-existent or out of date. ArkViz automates the creation and maintenance of software architecture diagrams, seamlessly and securely.

AI agents thrive when they have high quality context, including for architecture diagrams. However not all diagram formats out there are optimized for AI. A diagram showing connected systems and components can be represented in essentially 3 ways:

  1. Unstructured drawing / canvas format (e.g. draw.io, Excalidraw, Figma, LucidChart)
  - Quick to get started but hard to maintain.
  - Good for brainstorming and whiteboarding
  - The format carries no semantic meaning and is basically represented by pixels, shapes, and vectors. Positioning is stored
  - Passing this to an AI agent as context has questionable value.
  1. Structured diagramming formats (e.g. Mermaid, d2lang)
  - A level of abstraction higher than a 2D drawing canvas
  - Good for quick isolated diagram, like on GitHub
  - Represented by nodes and edges, nested hierarchies. Layout is generated
  - Better for AI agents, although they have to piece together the context, and might get confused by an inconsistency like duplicate naming
  1. Structured modeling formats (e.g. Structurizr, LikeC4, IcePanel).
  - These have an underlying data model and single source of truth
  - High quality documentation, but hard to create and maintain
  - Represented by entities and relationships. Layout can be richer, like zooming into a component and seeing what's inside
  - Best for AI agents, since they have a single source of truth context

ArkViz currently uses LikeC4 for its diagram format, a modern, popular, and open-source architecture-as-code format that supports interactive diagrams. AI agents will be more trained with this than a proprietary format. Why not roll my own solution with something like Claude Code? While Claude Code is a powerful AI coding harness, ArkViz provides seamless features you'd rather not spend time building like hosting diagrams, access control, scheduling automation, handling errors and retries, logs, connectors to GitHub. Additionally, ArkViz has battle-tested, optimized multi-pass prompts, validators, and guardrails for the specific purpose of AI generated architecture diagrams.

Take a look here: https://arkviz.io! Feedback greatly appreciated on which features you care about, and whether you think the value proposition is there.

Comments URL: https://news.ycombinator.com/item?id=50580994

Points: 1

── more in #ai-agents 4 stories · sorted by recency
── more on @arkviz 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/show-hn-arkviz-ai-ge…] indexed:0 read:3min 2026-11-20 · —