A spatial operating system for AI agents.
Agentarium is not a dashboard with 3D wallpaper. It is an event-driven world where agents can think, build, debate, inspect, and leave behind artifacts that matter.
Most agent products disappear into:
- chat logs
- terminal scrollback
- dashboards with no spatial meaning
- one-shot outputs with no replay value
Agentarium makes agent work legible through:
- spatial semantics
- observable agent roles
- artifact-first workflows
- human approvals at the right moments
- replayable sessions that can become content
This repository is the main product repo and canonical home for the codebase. Right now it contains the bootstrap structure, contributor scaffolding, and links to the public docs/spec work.
- Main code:
vultrdev/agentarium - Public docs/specs:
vultrdev/agentarium-docs - Release notes / upgrade history:
vultrdev/agentarium-releases
See REPOSITORIES.md for ownership and source-of-truth rules.
Core identity
- spatial operating system for AI agents
- browser-first
- event-driven
- inspectable
- artifact-first
- replay-first
MVP boundary
- one world
- one main room (Forge)
- three visible agents
- one event pipeline
- one artifact system
- one replayable session timeline
agentarium/
├── apps/
│ └── web/
├── packages/
│ └── contracts/
├── scripts/
├── docs/
└── .github/
The technical direction is intentionally simple:
- Next.js + React + TypeScript
- Three.js / React Three Fiber for the world
- normalized runtime events
- projected world state
- adapters for different agent runtimes
Simulation videos are coming soon.
Planned media includes:
- Forge room walkthroughs
- scripted session replays
- agent collaboration clips
- approval flow demos
- artifact lifecycle showcases
Media planning lives in the docs repo so the public narrative stays tidy.
- public docs index:
agentarium-docs/docs/README.md - product thesis:
01-product-thesis.md - world/runtime spec:
02-world-runtime-spec.md - MVP implementation plan:
03-mvp-implementation-plan.md
We want contributors from day one, but not chaos from day one.
If you want to help:
- read the docs repo first
- respect the runtime/world/command contracts
- keep changes scoped and legible
- prefer extension over mutation
- document architectural changes properly
Start with CONTRIBUTING.md.
- human-readable changes live in
CHANGELOG.md - canonical public release notes and upgrade guides live in agentarium-releases
- versioning follows Semantic Versioning
- changelog style follows Keep a Changelog
MIT