Beautiflow turns Mermaid source into polished, validated output. Use the CLI directly, keep a local preview open, or let the Agent Skill guide edits in Pi, Claude Code, or Codex.
architecture-beta group account(cloud)[AWS account] group vpc(cloud)[VPC · two AZs] in account group eks(cloud)[Amazon EKS] in vpc service users(lucide:users)[Public users] service route53(logos:aws-route53)[Route 53] in account service waf(logos:aws-waf)[WAF] in account service api(logos:aws-eks)[API] in eks service database(logos:aws-aurora)[Aurora] in vpc users:R --> L:route53 route53:R --> L:waf waf:B --> T:api api:R --> L:database
A diagram can render successfully and still be hard to follow. Beautiflow checks geometry, hierarchy, routing, and reproducibility before accepting a layout.
Keep Mermaid as the source of truth. Store layout separately, edit by meaning, audit the final geometry, and export diagrams ready for docs or slides.
Use one command for routine polishing. When a diagram needs a precise correction, semantic actions can change its layout or structure without editing pixel coordinates.
Read the Mermaid family, nodes, connections, groups, labels, and saved layout preferences.
Generate a limited set of candidates while preserving pinned nodes, primary paths, and semantic roles.
Reject layouts with overlaps, blocked arrows, excessive crossings, unreadable labels, or quality regressions.
Write SVG or PNG, render terminal output where supported, or keep a read-only browser preview open.
The same source and saved layout produce the same result. Parsing, routing, validation, transformation, and rendering stay local and deterministic.
$ beautiflow polish architecture.mmd $ beautiflow server architecture.mmd
Conversation, visual review, and intent stay in your coding harness. The portable skill tells the agent how to use the CLI and when a valid result is good enough to stop.
$ beautiflow install-skill Then ask: Use Beautiflow to polish architecture.mmd.
Every example includes the Mermaid source, the rendered result, the exact command that produced it, a live-preview command, and an optional prompt for your coding agent.
A 26-service architecture with explicit ports, recognizable service icons, nested account, VPC, and subnet boundaries, plus a stronger visual path for primary traffic.
architecture-beta group account(cloud)[AWS account] group vpc(cloud)[VPC · two AZs] in account group eksGroup(cloud)[Amazon EKS] in vpc service route53(logos:aws-route53)[Route 53] service cloudfront(logos:aws-cloudfront)[CloudFront] service waf(logos:aws-waf)[WAF] service controller(logos:aws-eks)[Controller]
A complete flowchart with a clear activation path and a payment-recovery loop that stays readable without competing with the successful path.
flowchart TD
signup([Trial signup]) --> profile[Create profile]
profile --> plan{Select a plan?}
plan -->|Plan selected| payment[Verify payment]
payment --> accepted{Payment accepted?}
accepted -->|No| retry[Request another method]
accepted -->|Yes| provision[Provision workspace]
Sequence, class, ER, and chart diagrams use dedicated renderers, so each family keeps the visual conventions that make it easy to read.
sequenceDiagram participant Customer participant Support participant Diagnostics participant Specialist Customer->>Support: Report intermittent failure Support->>Diagnostics: Collect recent traces Specialist-->>Support: Recommend configuration fix Support-->>Customer: Explain and apply fix
Each standalone release includes the complete rendering pipeline, with no Bun installation required. The macOS and Linux binaries are built and tested on their target operating systems.
Download the binary, add it to your PATH, and run it on a Mermaid file.
macOS Apple silicon · Linux x86-64
curl -fsSL https://beautiflow.cc/install.sh | bash
Start with one Mermaid file. Beautiflow preserves its meaning while making the layout clear, validated, and reproducible.