What it actually does
Pascal is a browser-based 3D building editor — floorplans, BIM, parametric design — built with React Three Fiber and WebGPU. That alone would put it outside most agent directories.
What earns it a place is the MCP service. Running `npx @pascal-app/cli editor` starts the editor *and* an authenticated MCP service in the background, picking collision-free loopback ports and keeping projects in a local SQLite database. You point an agent at `pascal mcp connect` and it can now manipulate architectural models directly.
That combination is unusual and instructive. It is a concrete example of a domain application exposing itself to agents through MCP rather than expecting an agent to click through its interface — the pattern most vertical software will eventually need to adopt.
MIT licensed, 21,000 stars, published as npm packages (`@pascal-app/core`, `viewer`, `cli`) so pieces can be embedded elsewhere.
Who it's for
- 01
Architecture, construction, and real-estate teams working with 3D building data
- 02
Product teams looking for a working reference on exposing an app to agents over MCP
- 03
Developers embedding a 3D viewer into their own property or design product
- 04
Anyone generating or modifying floorplans programmatically
Where it earns its keep
- Letting an agent generate floorplan variants from a written brief
- Embedding the viewer into a listing or property management product
- Studying how a complex GUI application exposes safe operations to an agent
- Producing consistent 3D visuals for proposals without a designer for each one
- Keeping architectural project data local in SQLite rather than a vendor cloud
Use it, or skip it
Reach for it when
- You work with building or spatial data and want agent access to it
- You need a self-hostable 3D editor rather than a subscription product
- You are designing your own MCP surface and want a real example to copy
- Embedding a viewer matters more than building one
Skip it when
- You need certified structural or regulatory-grade CAD output
- Your users are on hardware without WebGPU support
- You have no 3D or spatial dimension to your product — this is genuinely domain-specific
- You need the ecosystem and file-format support of established CAD tooling
10 automations
Ideas, not tutorials. Each one is work a team does by hand today.
- 01Operations
Brief to floorplan variants
An agent reads a client brief and generates several floorplan options for a human to react to in the first meeting.
- 02Marketing
Listing visualisation pipeline
Generate consistent 3D views for every property listing automatically instead of commissioning each one.
- 03Sales
Proposal asset generation
Produce branded 3D visuals for sales proposals on demand, so a designer is not in the critical path.
- 04Engineering
MCP surface reference
Use its authenticated MCP service as the pattern for exposing your own product's operations to agents safely.
- 05Operations
Spatial data validation
Have an agent check submitted plans against dimensional rules and flag violations before human review.
- 06Operations
Client revision loop
Capture client feedback in plain language and have the agent apply the change, with a person approving the result.
- 07Marketing
Portfolio rendering batch
Re-render an entire project portfolio to a new visual standard overnight rather than one at a time.
- 08Engineering
Embedded viewer in your product
Drop the viewer package into your own property or design app instead of building 3D rendering yourself.
- 09Operations
Local project archive
Keep all project data in the local SQLite store and back it up on your own schedule, with no vendor dependency.
- 10Operations
Design brief digest
Summarise the geometric changes between two project versions into a plain-English changelog for the client.
Want one of these running by Friday?
LimeDock builds these as real workflows inside your stack — deployed to your cloud, wired into your Slack and CRM, with the code in your repo. You pay a build fee and your own API keys, nothing else.
Source
Repository stats were read from the GitHub API and reflect the last time we refreshed this entry. The editorial breakdown above is LimeDock’s own analysis — we are not affiliated with pascalorg.
https://github.com/pascalorg/editor