Resources
Back to studioCase Study
Case study

Hypernovum

A spatial command center for AI agents. Free Obsidian plugin, standalone Pro app.

open sourceMIT
$ git clone
github.com/Pardesco/hypernovum
+ standalone Pro tier for supporters
The setup

AI has brains (LLMs) and hands (agents), but developers still orchestrate them through linear chat windows — scrolling walls of text, switching tabs, losing context, forgetting which project they were in. There is no spatial interface for managing AI work across many projects. Hypernovum is the cockpit.

Design decisions

Three choices that mattered.

  1. 01

    Spatial UI as the cockpit.

    Projects become 3D buildings in a city; agents launch from buildings rather than from another chat tab. Right-click a building to open its terminal, fire an agent, see git status. The foundational reframe is to stop watching terminals and start commanding a place — 3D as utility, not novelty.

  2. 02

    Ambient state via Git Weather and memory glow.

    Buildings pulse warm on recent commits, cool on staleness, and glow brighter as agent memory deepens for that project. Status is felt peripherally, not read on a chart. Trades dashboard-style reporting (Jellyfish, LinearB) for perception-level awareness — you learn the shape of your portfolio at a glance, without reading any metrics.

  3. 03

    Persistent agent memory plus agent-agnostic architecture.

    AI sessions normally start cold; agents re-learn the codebase every time. Engram persists per-project context inside each building — architectural decisions, error patterns, conventions — so launching an agent inherits the project's memory automatically. Combined with first-class support for any CLI agent (Claude Code, Gemini, Codex, Aider, Cursor), this solves the cold-start problem without trapping the user in one AI ecosystem.

AI & tool choices

AI is deep inside the product: Engram memory layer, MCP server that exposes city state to MCP-compatible clients (Claude Code, Cursor, Windsurf, VS Code, Claude Desktop), and Tandem Terminal for running two agents in parallel on a single codebase. The product is, deliberately, an AI UX thesis: command spaces, not chats; ambient state, not dashboards; portable memory, not vendor lock-in. Built with Claude Code and Gemini CLI as collaborators.

Result

Shipped free OSS Obsidian plugin (AGPL-3.0) plus standalone Pro desktop app for Windows and macOS. One-time pricing — $15 lifetime with a 30-day money-back guarantee. Real adopter testimonials live on the product page. MCP integration shipped.