Free house plan generator: describe a home, download the whole plan set

Type what you want, like "3 bed 2 bath farmhouse, 1,400 sq ft, porch". Open House Plans lays out the rooms, sizes the headers and joists, checks egress and light, takes off the materials, estimates the cost, and gives you a floor plan, elevations, a 3D model, CAD and BIM files. Every design is public domain (CC0): use it, change it, build from it. Nothing is uploaded.

Describe your house

All settings (the sentence fills these in; change any of them)

Designing the example...

Ready-made free house plans

Twelve designs from the generator, each a brief you can open, change and download. All public domain.

For AI agents and developers

The engine is dependency-free JavaScript. An agent can design houses from instructions without a browser:

# a summary as JSON (rooms, cost, checks, runners-up)
node platform/agents/tools.mjs call open-house-plans design --input '{"text":"3 bed craftsman, 1500 sq ft"}'
# the whole plan set written to a folder, plus a zip
node ventures/open-house-plans/cli.mjs design --beds 4 --baths 2.5 --garage 2 --style ranch --out my-house --zip
# or any MCP client: node platform/agents/tools.mjs mcp  (tools: search_tools, describe_tool, call_tool)

A brief is JSON; every field is optional: beds (0-5), baths (1-3.5), area (sq ft), style (farmhouse, craftsman, modern, ranch, cottage, minimal), priority (budget, balanced, beauty), garage (0-2), porch, foundation (slab, crawl), frame (advanced, conventional), accessible, split, lot ({width, depth, front, side, rear}), maxWidth, maxDepth, snow, wind, finish, region, ceiling, variant, or text in plain English. The same brief always gives the same house. The agent guide is ventures/open-house-plans/AGENTS.md in the source code.

How it works

The plan is a rectangle on a 6 inch grid, laid out as zones along the street: an optional garage, a great room (living, dining and kitchen, open plan), a primary suite on its own or at the end of the wing, and a bedroom wing with a hall and a row of rooms on each side. Small homes stack the bedroom and bath beside the great room instead. The search tries thousands of footprints and arrangements and scores each one on cost, the area you asked for, room sizes and proportions, plumbing runs, hall area, the front facade's symmetry and rhythm, and daylight, weighted by what you optimise for.

Then it places doors and windows (egress casements in every bedroom), sizes headers, studs and floor framing with the NDS checks from the open span tool and framing study, takes off the materials, and estimates the cost from NAHB's 2024 survey scaled by the design's own geometry. One 3D model drives the views, the elevations and the IFC and OBJ files, so they always agree.

Are these house plans really free to build from?

Yes. Every plan set the generator makes is dedicated to the public domain under CC0 1.0: you can use, change, sell and build from it with no permission and no credit. CC0 is not a warranty. A real house still needs a licensed review or the local code's prescriptive path, site and soils work, and a permit.

Can I build directly from these drawings?

No. They are design-development drawings generated by software and are marked not for construction. Take the files to a licensed architect, engineer or a builder working under your local code. The IFC and DXF files are meant to save them time.

What files do I get?

A presentation sheet, a dimensioned floor plan and four elevations as SVG; the floor plan as DXF for CAD; a 3D model as IFC4 (BIM, with named rooms) and OBJ; a bill of materials, window and door schedule, room schedule and cost breakdown as CSV; the design as JSON; a README and the CC0 licence.

How is the construction cost estimated?

It is anchored on NAHB's 2024 Cost of Constructing a Home survey (about $162 per square foot for the average 2,647 sq ft house) and scaled stage by stage with this design's own geometry: wall and roof area, footprint, perimeter, openings and plumbing fixtures. The split of each stage over those drivers is an assumption written out in the code. It is a model, not a quote.

How does the generator choose a layout?

It enumerates thousands of layouts on a 6 inch grid (footprint, zones, hall position, which rooms sit on which side, kitchen size and side) and scores each on cost, how close it is to the area you asked for, room sizes and proportions, plumbing runs, hall area, facade symmetry and daylight, weighted by your priority. The same brief always gives the same house.

Can an AI agent use it?

Yes. The engine is plain JavaScript with no dependencies. An agent can call design() with a JSON brief through the repository's command line or MCP server, or run cli.mjs to write a whole plan set to a folder.

Does it check building codes?

It runs checks paraphrased from the 2021 International Residential Code by section number (bedroom size, ceiling height, glazing, emergency escape windows, hall width, egress door) plus NDS checks on headers, studs and floor framing. Passing them is not code compliance: the code text was not copied or checked line by line, and your jurisdiction may differ.

What it does not do yet

Single-storey rectangles only (no second floor, stairs, L-shapes or basements); no site plan, foundation design for your soil, braced-wall design, truss design, mechanical, electrical or plumbing drawings, or energy-code compliance.