Skip to content
Get packs

Guide · Open source

How to use Pascal Editor: run it locally, self-host and connect AI agents

Updated · written with AI from the project's own documentation and checked against our catalogue data

Pascal Editor is an open-source, local-first 3D building editor that runs in the browser or from a command-line tool, and it can be connected to AI agents through the Model Context Protocol (MCP). This guide is for people who want to model a building on their own machine, developers who want an agent to work on a building scene, and teams that want to self-host the editor. By the end you will know how to start it locally, where projects are stored, how to self-host it with the published Docker image, and the limits to keep in mind.

Pascal Editorpascalorg/editor · ★ 25.3k · MIT

Quick start

  1. 1Check your Node.js version. Confirm you have Node.js 22.23.4+ on 22.x, 24.22.1+ on 24.x, or 26.12.1+ before you begin.
  2. 2Start the editor. Run npx @pascal-app/cli editor in a terminal. The CLI starts the editor and an MCP service in the background.
  3. 3Find your projects. Projects are kept in ~/.pascal/data/pascal.db on your machine.
  4. 4Install the pascal command (optional). Run npm install --global @pascal-app/cli to use the pascal command repeatedly.
  5. 5Connect an agent. Configure your AI agent to launch pascal mcp connect, and keep only one active agent client per local CLI service.
  6. 6Self-host with Docker (optional). Run the ghcr.io/pascalorg/pascal-editor image with a volume for /data, an admin password and your PASCAL_PUBLIC_URL.

Check the requirements before you start

Pascal Editor is built with Node.js tooling, and the README lists the Node.js versions that work: 22.23.4 or newer on the 22.x line, 24.22.1 or newer on the 24.x line, or 26.12.1 or newer. Setup involves Node.js, and the self-hosting route also involves Docker, so people without experience in either should expect a learning curve.

  • Node.js 22.23.4+ on 22.x, 24.22.1+ on 24.x, or 26.12.1+
  • Docker, only if you plan to self-host the container
  • The pascal command is optional; npx runs the editor without installing it

Run the editor locally with npx

The quickest route needs no cloning of the repository. Run the command below in a terminal. The CLI starts the editor and an authenticated MCP service in the background, selects loopback ports that do not collide, and keeps projects in a local database.

Projects are stored in ~/.pascal/data/pascal.db. The web editor runtime is downloaded once per version on the first command that starts the editor and is verified against a digest published inside the npm package.

npx @pascal-app/cli editor

Install the pascal command globally

If you want to run the CLI repeatedly, install the pascal command globally. The README gives this install command. After that, you can start the editor with pascal editor and use the other CLI commands.

The README points to the full local editor documentation for pnpm and Bun commands, project management, MCP setup, updates, storage paths and troubleshooting, so check that page for anything beyond the basics.

npm install --global @pascal-app/cli

Connect an AI agent over MCP

Pascal Editor exposes its scene to AI agents through MCP. Configure your agent to launch pascal mcp connect. This connector does not need the editor process to be running, and the README says the local connector needs no Pascal account or API key and does not upload projects automatically.

Use only one active agent client per local CLI service. The README notes that the standalone local HTTP runtime shares active scene state between clients, so if you need independent concurrent work, use separate PASCAL_HOME directories and service processes.

  • Configure the agent to launch: pascal mcp connect
  • Keep one active agent client per local CLI service
  • For independent parallel work, use separate PASCAL_HOME directories and service processes

Self-host the editor with Docker

The editor ships as one container image, ghcr.io/pascalorg/pascal-editor. Its tags match the npm package versions, and the latest tag follows stable releases. Projects live in SQLite under /data, one admin password protects the instance, and PASCAL_PUBLIC_URL is read at runtime so the same image can serve any domain.

The example below maps port 3000, stores data in a named volume and sets the admin password and public URL. Replace change-me and the example domain with your own values. The README points to SETUP.md for the compose file, the full list of variables, agent access and public embeds.

  • Use a strong value for PASCAL_ADMIN_PASSWORD instead of change-me
  • Set PASCAL_PUBLIC_URL to the domain you serve the editor on
docker run -d -p 3000:3000 -v pascal-data:/data \ -e PASCAL_ADMIN_PASSWORD=change-me \ -e PASCAL_PUBLIC_URL=https://pascal.example.com \ ghcr.io/pascalorg/pascal-editor:latest

Understand the agent skills and their limits

The repository provides agent skills that can be installed with skills.sh. The README names two: pascal-3d, which covers safe local or hosted MCP setup and verified scene work, and furniture-fit, which produces a bounded, evidence-based footprint assessment. The skills check the connected MCP tool schemas before using optional fields, so an older installed release may give a narrower result.

The README also says the live demo is a preview of the direction of the next Pascal editor, not part of the released open-source editor, so do not assume every feature shown there is available to you.

  • Install skills with npx skills add pascalorg/editor, choosing pascal-3d and furniture-fit
  • The Claude Code plugin needs the pascal CLI installed and on the PATH used to launch Claude Code
  • Remove duplicate Pascal MCP entries so only one active connection exists per local service
npx skills add pascalorg/editor \ --skill pascal-3d \ --skill furniture-fit

What people use it for

  • Sketching or modeling a home or building in 3D on your own computer
  • Letting an AI agent make verified changes to a 3D building scene over MCP
  • Running a shared, self-hosted building editor for a small team with the published Docker image
  • Exploring the structure and layout of a building as a hobbyist interested in BIM-style tools

Frequently asked questions

Is Pascal Editor free to use?

The project is released under the MIT licence, as stated in the README. Check the LICENSE file in the repository for the full terms before you use it in a commercial setting.

How do I install Pascal Editor?

The quickest way is to run npx @pascal-app/cli editor, which needs no cloning. For a reusable command, run npm install --global @pascal-app/cli to install the pascal command.

What are the requirements for Pascal Editor?

You need Node.js 22.23.4+ on 22.x, 24.22.1+ on 24.x, or 26.12.1+. If you self-host, you also need Docker.

Where are my Pascal projects saved?

When you run the local CLI, projects are kept in ~/.pascal/data/pascal.db. Self-hosted instances store projects in SQLite under /data inside the container.

Can several AI agents edit the same local scene at once?

The README says to use one active agent client per local CLI service. For independent concurrent work, use separate PASCAL_HOME directories and service processes.

Is there an alternative to Pascal Editor?

The source does not name comparable projects, so this guide cannot make a comparison. Evaluate it against your own needs for 3D modelling and MCP support.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Pascal Editor suits people who want to model a building in 3D on their own machine, developers who want to connect an AI agent to a scene over MCP, and teams comfortable self-hosting with Docker. It is not a finished, production-certified architectural or engineering tool, and it is not the right choice if you need several agents editing the same local scene at once or you have no experience with Node.js or Docker.

Install options & downloads