Booboo fuses wiring, knowledge, memory, agents and crons into a single rooted 3D graph — queryable by REST or MCP, rendered at 60fps in your browser.
Proven on a 4,507-node production brain · built from Supabase by config alone · privacy-walled
Booboo is a ~1 KB JSON contract with adapters that feed it and consumers that render, serve and query it. Emit the JSON once — get the viewer, the API and the MCP server for free.
Postgres, Supabase, plain JSON — config-driven, no code. Weird source? A custom adapter is ~50 lines, not a fork.
A rooted snapshot of everything — nodes, links, layers. Privacy walls applied before emit: sealed namespaces are never serialised.
3D viewer in the browser · REST API · MCP server your agents boot from · desktop wallpaper. All from the same snapshot.
Knowledge graphs, agent flowcharts, memory stores, trace viewers — each is a fragment. Booboo is the fusion, rooted at a single point.
An instanced GPU field with tier-LOD renders 1,000,000 nodes at 60fps — in a browser tab, no install. booboo view opens any snapshot.
booboo serve gives you /graph /search /neighbors /path — your whole system, queryable over HTTP in one command.
Point Claude, Cursor or Claude Code at booboo mcp and your agents boot from the brain — search, node lookup, pathfinding as tools.
One booboo.config.yaml maps tables to layers. The 4,507-node proof brain was assembled from a live Supabase with zero custom code.
Sealed namespaces are excluded before the snapshot exists. What's walled never reaches the JSON — not hidden in the viewer, absent from the artifact.
The engine is free forever — fork it, adapt it, ship it. Fractional HQ sells the done-for-you layer on top, never the core.
Scaffold a project, point booboo.config.yaml at your Postgres or JSON, and build. The same snapshot powers the API, the MCP server and the 3D view.
No hosted service, no account, no telemetry. It runs on your machine, against your data.
$ npx create-booboo my-brain $ cd my-brain && npm install $ npm run build # config → brain.json ✓ 4,507 nodes · 21 links · walls applied $ npm run view # 3D brain in your browser $ npm run mcp # agents boot from it
50,000 nodes, four layers, live in a browser tab — one command, no data wired. Yours looks like this on day zero.
Unretouched screenshot · booboo view --demo --nodes 50000 · zero console errors · try it yourself in one command
Good tools, all of them — for their layer. None of them is a map of the system.
| Tool | Whole-system view | REST API | MCP — agents query it | 3D at 1M nodes | Privacy walls |
|---|---|---|---|---|---|
| Booboo — the operational brain | ✓ | ✓ | ✓ | ✓ | ✓ |
| Graph viewers (3d-force-graph) | render only — nothing feeds them | — | — | ✓ | — |
| Note graphs (Obsidian, Logseq) | your notes, not your system | — | plugins | — | — |
| Agent frameworks (LangGraph, traces) | flows & runs — one slice | ✓ | partial | — | — |
| Memory stores (Graphiti, Cognee) | memory — another slice | ✓ | ✓ | — | — |
The code is MIT and always will be — nothing ever hides behind a paywall. You pay only to skip the work: key in hand for £29, or our hands entirely.
npx create-booboo scaffoldYou're not buying features — every feature is free. You're buying the deployed outcome: your agent does the whole setup while you answer five questions.
One command stands between your stack and a brain you can query, render and boot agents from.