openspec-ui (ToruAI) — Summary

openspec-ui is a standalone multi-repository dashboard for tracking OpenSpec spec-driven development activity across multiple projects, delivered as a cross-platform binary (Rust/Axum backend + React/TypeScript frontend) that reads openspec/ directory structures and presents them as a real-time Kanban board with live file-watching.

Problem it solves: When using AI coding assistants across multiple projects simultaneously, there is no central visibility into which features are in progress, what the AI finished, and what ideas are waiting — openspec-ui aggregates multiple openspec/ directories into a single dashboard so developers can capture ideas on mobile, track agent progress in real time, and manage the full idea→archived lifecycle from one interface.

Distinctive trait: The only framework in this batch that is a pure read/visualize tool with no AI interaction whatsoever. It reads OpenSpec file state and displays it; it does not prompt AI, send commands, or drive workflows. Distributed as a native binary (Rust/Axum backend, React frontend) or Docker image — the most production-deployment-oriented entry in the batch.

Target audience: Developers running multiple AI-assisted projects in parallel who want mobile-accessible, real-time visibility into what their AI agents are doing.

Production-readiness: v0.1.0, 18 stars, last commit January 2026. Available as pre-built binary for macOS/Linux/Windows from GitHub Releases.

Relationship to seeds: Does not fit any of the 5 seed archetypes cleanly. The closest is claude-conductor (Archetype 4 — markdown scaffold, zero primitives) in that both treat the project's markdown files as the primary artifact and provide no AI interaction. But claude-conductor focuses on error ledgers and CLAUDE.md; openspec-ui focuses purely on multi-repo visualization. It is genuinely novel in the corpus: a passive dashboard observer with no AI primitives at all.

openspec-ui — Overview

Origin

Created by ToruAI (GitHub organization). The README opens with a distinctive "Why" section:

"AI coding assistants are powerful, but when you're working across multiple projects, it's hard to keep track of what's happening where. Which features are in progress? What did the agent finish yesterday? What ideas are waiting to be developed?"

Philosophy

The core philosophy is "mission control" — not a tool for driving AI agents, but for observing them:

"OpenSpec UI is the mission control — the actual spec-driven development happens through OpenSpec and your AI coding assistant (Cursor, Claude Code, etc.)."

The workflow model:

💡 Idea  →  📋 Proposal  →  ⚡ Implementation  →  ✅ Done
   ↑            ↑                  ↑
  You      AI Agent           AI Agent

• You capture ideas in the UI
• AI agents expand ideas into proposals and implement them
• UI shows progress passively

Key design decisions

1. No AI integration: The dashboard does not send prompts or interact with any AI. It is read-only with one exception: the "capture idea" feature writes a quick note to the openspec/ideas/ directory.
2. Multi-repo focus: Uniquely designed from the start for multiple repositories. The config file lists multiple openspec/ directory paths.
3. Mobile-first: "Works great on phone and tablet" — the UI uses responsive design so developers can check progress from mobile.
4. Real-time via SSE: Uses Server-Sent Events (not WebSocket) for file change notifications.
5. Binary distribution: Primary install path is a pre-compiled binary, not npm or Docker-first.

openspec-ui — Architecture

Distribution

• Type: Standalone binary (Rust backend + React frontend)
• Also available: Docker image, build from source

Tech stack

Install methods

# Option 1: Download binary (recommended)
# Download from GitHub Releases:
# - openspec-ui-v0.1.0-darwin-aarch64.zip  (macOS Apple Silicon)
# - openspec-ui-v0.1.0-linux-x86_64.zip    (Linux x86_64)
# - openspec-ui-v0.1.0-windows-x86_64.zip  (Windows x86_64)
unzip openspec-ui-*.zip
./openspec-ui --config openspec-ui.json

# Option 2: Docker
docker build -t openspec-ui .
docker run -p 3000:3000 \
  -v /path/to/your/repos:/repos \
  -v /path/to/openspec-ui.json:/app/openspec-ui.json \
  openspec-ui

# Option 3: Build from source
cd frontend && npm ci && npm run build && cd ..
cd backend && cargo build --release && cd ..
./backend/target/release/openspec-ui --config openspec-ui.json

Configuration file

openspec-ui.json:

{
  "sources": [
    { "name": "my-project", "path": "/path/to/my-project/openspec" },
    { "name": "another-repo", "path": "/path/to/another-repo/openspec" }
  ],
  "port": 3000
}

Repository structure

openspec-ui/
├── backend/
│   ├── Cargo.toml      ← Rust project manifest
│   └── src/            ← Axum server code
├── frontend/
│   ├── package.json    ← React + Tailwind + shadcn/ui
│   ├── src/            ← React components
│   └── vite.config.ts
├── openspec/           ← Sample openspec data (for dev)
├── Dockerfile
├── build_n_run.sh      ← Quick build+run script
├── openspec-ui.example.json
└── package.json        ← Workspace root (dev scripts)

Target AI tools

None — the dashboard does not interact with any AI tool. It reads openspec/ directories written by any tool.

Reads OpenSpec directory structure

openspec/
├── specs/          ← Source of truth specs
├── changes/        ← Active changes (proposals + tasks)
└── ideas/          ← Quick ideas

openspec-ui — Components

Dashboard pages/views

Features

• Kanban Board: Ideas → Todo → In Progress → Done → Archived columns
• Multi-repo visibility: Monitor multiple OpenSpec repositories from one dashboard
• Real-time updates: Auto-refreshes when files change (SSE)
• Idea capture: Quick-capture thoughts that AI agents can later expand
• Mobile-first: Responsive design works on phone and tablet
• Light/Dark Theme: Toggle

Commands

Count: 0 — No slash commands, no CLI subcommands, no SKILL.md files.

Skills

Count: 0

Subagents

Count: 0

Hooks

Count: 0

MCP Servers

Count: 0

Scripts

Templates

Count: 1

openspec-ui — Uniqueness and Positioning

differs_from_seeds

None of the 11 seeds are pure visualization dashboards. The closest is claude-conductor (Archetype 4 — markdown scaffold, zero primitives) in that both treat the project's markdown files as first-class artifacts and neither drives AI directly. But claude-conductor generates and manages CLAUDE.md/ERRORS.md/TASKS.md; openspec-ui only reads and visualizes. openspec-ui is genuinely outside the 5-archetype taxonomy: it is the only "passive observer" in the corpus — an operations dashboard for AI-generated work rather than a framework that shapes AI behavior. Compared to openspec-mcp (which also ships a web dashboard), openspec-ui has zero AI interaction and focuses on multi-repo aggregation rather than single-project tool-calling.

Positioning

• Niche: "Mission control dashboard for multi-repo OpenSpec work"
• Primary differentiator: Only pure visualization tool in the batch; no AI interaction
• Secondary differentiator: Multi-repo aggregation — designed from the start to watch N projects simultaneously
• Distribution differentiator: Native binary (Rust) + Docker — the most deployment-ready tool in the batch

Observable failure modes

1. No authentication: Running the server on a non-loopback interface exposes all openspec content to anyone on the network
2. Read-only except ideas: The only write operation is idea capture; all other state changes must go through the OpenSpec CLI or AI agents
3. Stale taxonomy: The Kanban column logic (mapping checkbox states to columns) may not match evolving OpenSpec state conventions
4. Single binary version: No hot-reload or rolling update; updating the binary requires restart
5. No notification system: No push alerts when an agent completes a task; users must actively check the dashboard

Explicit antipatterns

None documented.