Pascal Editor

Active
GitHub TypeScript MIT

Description

Open-source local-first 3D building editor built with React Three Fiber and WebGPU, connecting AI agents through MCP to read and edit scenes from the browser or CLI.

Key Features

  • Local-first runtime — a single npx command starts a persistent local service with projects kept in ~/.pascal
  • MCP agent access — a built-in MCP server exposes scene tools, resources, and prompts to Claude Code, Codex, and other hosts
  • WebGPU 3D rendering — a React Three Fiber runtime with stacked, exploded, and solo level viewing modes
  • Node-based scenes — Site, Building, Level, Wall, and Zone nodes live in a flat dictionary with Zod schemas and undo/redo
  • On-demand geometry systems — WallSystem, SlabSystem, and friends recompute only dirty nodes, with CSG cutouts for doors and windows
  • Plugin extensibility — new node kinds and side panels ship as plugins using the same manifest as the built-ins

Use Cases

💡 Direct an AI agent in natural language to lay out floors, walls, and furniture while MCP syncs the scene.
💡 Embed the @pascal-app/viewer package in a web app to add 3D floor-plan or spatial visualization to your product.
💡 Write custom node plugins to extend the editor with procedural assets such as trees and vegetation.
💡 Keep housing layout projects on your own machine thanks to local-first storage.

Strengths & Limitations

Strengths

  • Actively maintained, recent updates
  • High community interest (23.0k stars)
  • Permissive open-source license (MIT)
  • Responsive to issues, low backlog

Quick Start

# Start the local editor (requires Node.js 22.13+)
npx @pascal-app/cli editor
# Configure the MCP connection for Claude Code
pascal mcp setup claude
# Install the official agent skills
npx skills add pascalorg/editor --skill pascal-3d --skill furniture-fit
# Develop from source
bun install
bun dev

Related Projects

Awesome LLM Skills

1.7k · Python
Normal C

A curated list of awesome LLM and AI Agent Skills, resources and tools for customising AI Agent workflows. Works with Claude Code, Codex, Gemini CLI and custom agents.

agenttoolsmcp +2
  • · Curated collection of LLM skills for Claude Code, Codex, Gemini CLI, and more
  • · Covers document processing (docx, pdf, pptx, xlsx), development tools, and research workflows
  • · Cross-platform compatibility with Anthropic, OpenAI, Google, and Alibaba AI tools

Continuous Claude

3.9k · Python
Stale C

Context management for Claude Code with hooks for state maintenance via ledgers and handoffs. Enables MCP execution without context pollution and agent orchestration with isolated context windows for long-running conversations.

mcpagentpython +3
  • · Persistent learning system with YAML handoffs that preserve context across sessions, solving Claude Code's compaction problem
  • · 109 natural-language-triggered skills with automatic activation via rule-based context injection hooks
  • · 32 specialized agents including scout (exploration), oracle (research), debug-agent, and kraken for parallelized task execution

GhidraMCP

10.0k · Java
Stale B

MCP server for Ghidra reverse engineering platform, enabling AI agents to autonomously perform binary analysis and vulnerability discovery.

mcpreverse-engineeringghidra +2
  • · MCP server for Ghidra enabling LLMs to autonomously perform binary reverse engineering
  • · Decompile and analyze binaries with automatic renaming of methods and data
  • · List all methods, classes, imports, and exports, exposing core Ghidra functionality to MCP clients