AI-native diagram tool — renders 28+ engines (Mermaid, PlantUML, D2, Vega-Lite, TikZ, Graphviz, WaveDrom, Bytefield, etc.) with annotations, multi-tile infinite canvas, and shared-workspace AI loop.
Use when the user references annotations or marks they've made on a diagram — phrases like "what did I tag?", "summarize my annotations", "explain the regions I marked", "what do my notes say?". Also triggers when the user asks an open-ended question about a diagram and there are unresolved annotations on the focused tile.
Use when the user has just added annotations to a diagram and is implicitly or explicitly asking for the diagram to be changed based on those marks — phrases like "fix this", "apply my notes", "make those changes", "address what I marked". Triggers on evolve/apply intent against an annotated tile.
Use whenever the user wants any kind of diagram, chart, flowchart, sequence diagram, ER diagram, architecture diagram, state machine, packet layout, signal timing, network topology, or visual representation of structure. Triggers on engine names (Mermaid, PlantUML, D2, Vega-Lite, TikZ, Graphviz, etc.), diagram-type words (flowchart, sequence, ER, state, class, packet, waveform, scaffold), and topics that benefit from visual structure (architectures, data flows, schemas, protocols).
Use when the user has saved diagrams in the project library and asks for review-style help — phrases like "look over my architecture", "any issues with this diagram?", "review what I have", "what would you change?". Triggers on review/critique intent against existing saved content.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
▶ Watch the 30-second demo in HD
AI-native diagram tool. 28+ rendering engines (Mermaid, PlantUML, D2, Vega-Lite, TikZ, Graphviz, WaveDrom, Bytefield, …) wrapped behind an MCP server, with annotations, infinite-canvas multi-tile workspace, and a marketing-surface workspace UI. Cycle 4 ships as a hosted service with optional self-host.
Visit https://prixmaviz.ailuxis.com — your workspace is created automatically and the URL is bookmarkable. Lose the URL, lose the workspace, so save it somewhere.
Supported platforms: macOS (Apple Silicon + Intel), Linux (x64 + arm64), Windows (x64). The plugin install itself is tiny (~3 KB launcher script). On your first diagram render, the launcher downloads the matching pre-built shim binary (~60–115 MB depending on platform) from this repo's GitHub Releases and caches it at ~/.cache/prixmaviz/bin/ (POSIX) or %LOCALAPPDATA%\prixmaviz\bin\ (Windows). Subsequent calls hit the cache — zero network.
claude plugins marketplace add MichaelDanCurtis/PrixmaViz
claude plugins install prixmaviz@prixmaviz
codex plugin marketplace add MichaelDanCurtis/PrixmaViz
Then enable the plugin in ~/.codex/config.toml:
[plugins."prixmaviz@prixmaviz"]
enabled = true
Restart Codex. Your next prompt that mentions diagrams will route through PrixmaViz.
The plugin uses https://prixmaviz.ailuxis.com by default. To point at your own self-hosted instance, set PRIXMAVIZ_REMOTE_URL in the cached .mcp.json (~/.claude/plugins/cache/.../<version>/.mcp.json or ~/.codex/.tmp/marketplaces/prixmaviz/src-tauri/resources/plugin/.mcp.json). To run a locally-built binary instead of the released one, set PRIXMAVIZ_MCP_BIN to its absolute path.
The AI will now use PrixmaViz whenever you ask for diagrams. Try:
Show me the TCP handshake.
git clone https://github.com/MichaelDanCurtis/PrixmaViz
cd PrixmaViz
cp .env.example .env
docker compose up -d
Open http://localhost:5180 — your workspace is ready. All 6 containers (server, postgres, kroki orchestrator, mermaid, bpmn, excalidraw sidecars) come up automatically. PlantUML is built into the orchestrator.
For production, set:
PRIXMAVIZ_PUBLIC_URL — the canonical URL diagrams share (e.g., https://prixmaviz.yourcorp.com)DATABASE_URL — your external Postgres (omit to use the bundled one)Anonymous workspaces expire after 1 hour of inactivity. Activity means any authenticated /api/* request — keeping a browser tab open with the workspace loaded is sufficient.
Pin a workspace indefinitely by toggling at least one diagram to public (🌐 icon on the tile header). Workspaces with any public diagram are exempt from TTL.
To change the policy, set in your .env:
PRIXMAVIZ_WORKSPACE_TTL_MINUTES=60 — how long before idle workspaces expire (set 0 to disable)PRIXMAVIZ_REAP_INTERVAL_MINUTES=5 — how often the reaper runs@prixmaviz/shim binary: thin MCP stdio→HTTP forwarder distributed via the CC plugin (5 platform builds: darwin-arm64/x64, linux-arm64/x64, windows-x64)KROKI_URL)See docs/superpowers/specs/2026-05-11-prixmaviz-cycle-4-design.md for the design rationale.
packages/
shared/ # TypeScript types shared between server, web, and shim
server/ # Bun HTTP+WS+MCP server, Postgres repos, hit-testers
web/ # React webview — InfiniteCanvas, Tile, public view, settings
shim/ # MCP stdio→HTTP forwarder (compiled to native binary)
src-tauri/ # Tauri 2 desktop wrapper (optional, for self-hosted installers)
docs/
superpowers/
specs/ # Cycle design specs
plans/ # Cycle implementation plans
npx claudepluginhub michaeldancurtis/prixmavizGraph-native memory layer for AI coding agents.
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.
Memory compression system for Claude Code - persist context across sessions
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Write feature specs, plan roadmaps, and synthesize user research faster. Keep stakeholders updated and stay ahead of the competitive landscape.
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns