Luke The Dev
155 lines
4.5 KiB
Markdown
155 lines
4.5 KiB
Markdown
# Claw3D Vision
|
|
|
|
Claw3D is an open-source 3D environment for visualizing and interacting with AI agents powered by OpenClaw.
|
|
|
|
The long-term goal of Claw3D is to build a living 3D world where AI agents and humans collaborate: a kind of digital city where agents operate, communicate, and perform tasks in a shared visual space.
|
|
|
|
OpenClaw acts as the intelligence and orchestration engine, while Claw3D provides the visual layer and interactive environment that makes agent activity understandable, inspectable, and collaborative.
|
|
|
|
This document explains the direction of the project and the guardrails guiding its development.
|
|
|
|
Project overview and developer documentation can be found in:
|
|
|
|
- [`README.md`](README.md)
|
|
- [`ROADMAP.md`](ROADMAP.md)
|
|
- [`CONTRIBUTING.md`](CONTRIBUTING.md)
|
|
|
|
## Why Claw3D Exists
|
|
|
|
AI systems are becoming increasingly capable, but their behavior is often invisible or difficult to understand.
|
|
|
|
Claw3D aims to solve this by providing a visual interface for AI systems, allowing people to:
|
|
|
|
- observe AI agents operating in real time
|
|
- understand system behavior visually
|
|
- collaborate with AI in shared environments
|
|
- debug and inspect complex agent interactions
|
|
|
|
The ultimate vision is a 3D city of AI agents, where:
|
|
|
|
- agents represent services, tasks, and workflows
|
|
- humans can explore, monitor, and interact with them
|
|
- systems become understandable through spatial interaction
|
|
|
|
## Relationship to OpenClaw
|
|
|
|
Claw3D is designed to work with OpenClaw, not replace it.
|
|
|
|
OpenClaw provides:
|
|
|
|
- agent orchestration
|
|
- tools and integrations
|
|
- communication channels
|
|
- task execution
|
|
- model provider integrations
|
|
|
|
Claw3D provides:
|
|
|
|
- visualization
|
|
- interaction
|
|
- spatial representation of agents and systems
|
|
- collaborative environments for humans and AI
|
|
|
|
In simple terms:
|
|
|
|
```text
|
|
OpenClaw -> intelligence and task execution
|
|
Claw3D -> visualization and interaction layer
|
|
```
|
|
|
|
Maintaining compatibility with OpenClaw is an important design goal.
|
|
|
|
Features that require breaking OpenClaw integration will generally not be accepted unless there is a strong architectural reason.
|
|
|
|
## Current Priorities
|
|
|
|
Claw3D is still in an early stage of development.
|
|
|
|
Current priorities include:
|
|
|
|
### Stability and Reliability
|
|
|
|
- bug fixes
|
|
- predictable rendering behavior
|
|
- improving the developer experience
|
|
|
|
### Core Architecture
|
|
|
|
- defining how agents map to visual entities
|
|
- building a scalable world model
|
|
- establishing a clean integration path with OpenClaw
|
|
|
|
### Developer Ergonomics
|
|
|
|
- clear APIs for extending the environment
|
|
- easy local setup
|
|
- straightforward contribution paths
|
|
|
|
### Visualization Primitives
|
|
|
|
- representing agents
|
|
- representing workflows
|
|
- representing system activity in spatial form
|
|
|
|
## Contribution Rules
|
|
|
|
To keep the project maintainable:
|
|
|
|
- One PR = one topic. Avoid bundling unrelated changes.
|
|
- Very large PRs may be declined or split into smaller pieces.
|
|
- Architectural changes should be discussed in issues before implementation.
|
|
- Contributors should respect the project's direction and scope.
|
|
|
|
Claw3D is still evolving quickly, so iteration is expected.
|
|
|
|
## Architecture Direction
|
|
|
|
Claw3D is designed as a visual layer on top of agent systems.
|
|
|
|
The system should remain:
|
|
|
|
- modular
|
|
- extensible
|
|
- easy to experiment with
|
|
|
|
The current stack focuses on:
|
|
|
|
- Three.js
|
|
- WebGL
|
|
- browser-based rendering
|
|
- integration with OpenClaw runtime systems
|
|
|
|
The goal is to keep the environment accessible to developers and contributors.
|
|
|
|
## What We Will Not Merge (For Now)
|
|
|
|
To maintain focus, the following types of contributions are generally avoided:
|
|
|
|
- features that break compatibility with OpenClaw
|
|
- major architectural rewrites without prior discussion
|
|
- replacing the rendering stack without strong technical justification
|
|
- heavy framework layers that reduce hackability
|
|
- extremely large PRs without prior coordination
|
|
- unrelated product experiments that do not advance the Claw3D vision
|
|
|
|
This list is a directional guardrail, not a permanent restriction.
|
|
|
|
Strong technical arguments or user demand may change these decisions.
|
|
|
|
## Long-Term Direction
|
|
|
|
The long-term vision for Claw3D is ambitious:
|
|
|
|
**A 3D city of AI agents.**
|
|
|
|
In this environment:
|
|
|
|
- AI agents operate as visible entities
|
|
- systems become spatially understandable
|
|
- humans can interact with agent systems in real time
|
|
- collaboration between humans and AI becomes natural
|
|
|
|
Instead of interacting with invisible systems through logs and dashboards, users will be able to walk through and interact with the systems themselves.
|
|
|
|
Claw3D is an early step toward that future.
|