5 sheetsto a checked model of 2,420 parts on 20 floors
Exterior (generated)Floor plan3D model
From site photos
8 photosto a 3D model and a scaled plan
Site photo3D model2D plan
Eight exterior and interior photos of a plant, generated from a reference model so the error could be measured. Your own AI rebuilt the hall over MCP with Laydyne’s tools, machines and racks included, and checked it against each photo from the same viewpoint; dimensions stay estimates until one is measured. Read how it was built from photos
In the film: Calder Yards, a fictional 18-storey building with offices, homes, shops and a car park, modelled from five drawings over MCP.
5 sheets
floor plan images, as a client would send them
2,420 parts
on 20 floors, traced from the drawings and checked with no findings
778 people
in the weekday morning peak: 704 office workers and 74 residents through the lobby, café and lifts
307 s → 6 s
average coffee wait after one edit: the café opened into the next unit, with four baristas instead of two
Fictional drawings and synthetic demand. Studio opens the building with the café extended; the drawn layout is saved as option A.Photos: Market Hall, Charleston, Historic American Buildings Survey (Library of Congress, HABS SC,10-CHAR,6; no known restrictions). First estimate from the photos alone, against the measured drawings: width +6%, length +11%, ceiling height −18%. After one correction against the photos (the model shown): width −4%, length +11%, ceiling height −15%, window width −31%. A model from photos is an estimate to check against a measurement.
Simulate how people, robots and goods move.
Five examples, each the engine’s own playback: an assembly plant, a fulfilment centre, a shopping centre, a supermarket and an airport. Change the layout, the fleet or the staffing, run the same demand again and compare the result.
“Model our site, run the busiest hour, and compare two options.”
Assembly plant: AMRs, forklifts and operators share the aisles. Compare fleet sizes and routes.Fulfilment centre: robots bring shelves to the pick stations. Size the fleet and the stations.Shopping centre: Saturday visitors move between the shops and the food court. Test routes, escalators and peak crowding.Supermarket: shoppers queue at the checkouts. Compare the tills open and the layout.Airport: travellers through check-in and security. Compare lanes and staffing.Assembly plant: AMRs, forklifts and operators share the aisles. Compare fleet sizes and routes.Synthetic examples. Modelled flows, not a forecast for a real site.
32example spaces to open and change, from a café to an estate
2D · 3Ddrawings, pictures and playback from one model
Freefor modelling, editing, route checks, and project files without usage-count limits.
Whatever the building. Whoever you are.
If it has a floor plan and people, goods or robots moving through it, you can try a change and compare the result.
Made for robot integrators, store and restaurant operators, facility managers, architects and interior designers, event organizers and property developers.
“Keep the shell. Compare focus, teamwork, private studios, a client forum and window-side work. Save every option, check the routes and export the plans.”
Synthetic 30 × 20 m office. All ten saved layouts and five earlier comparisons remain. The five new proposals have geometry and route checks; their operating performance is not yet simulated.
A proposal you can hand over.
Follow a furniture reference from the plan to its quantity and part IDs. Open the actual A3 sheets and searchable proposal report.
A3 plans at 1:100Dimensions and furniture referencesQuantities tied to the modelSizes, counts and every part ID
Averages of 10 runs on the airport example, 90 passengers an hour. Across the whole journey, the average passenger waits 111 s less (95% interval 78 to 143 s). Synthetic example: modelled waiting, not a forecast for a real site.
Your workflow. Built through conversation.
Go beyond the samples. Ask AI to combine movement, tasks, arrivals, queues, and shared resources into a workflow that fits your operation.
“Add an inspection step after packing. Send 10% of orders back for rework, then compare one inspector with two.”
AI can compose custom scenarios from the supported process steps. The 2D/3D view lets you inspect its work, with manual editing available whenever you need it.
Build for free with your own AI.
Sign up free and connect Codex, Claude Code, or another MCP client. Create 2D/3D spaces, check routes, and prepare custom workflows. Ask your agent to export and restore project JSON using files you keep yourself.
Compare simulated operations under the same assumptions. Review queues, waits, and resource use, then export the results with their assumptions and limits. Use your own AI or the built-in assistant to refine the inputs.
Feature availability and usage limits are shown in Account after sign-up.
AI builds the model.
Engines do the maths.
AI changes the spatial model through shared tools. Geometry checks and dedicated engines calculate supported routes, cleaning, transport, and workflow measures. Keep the inputs, model version, and limits beside every result.
Export the starting project and reuse the same units and seed.
Take the work with you
Export your scene and conversation as a project file.
Your AI. Your space. Your files.
Build and edit 2D/3D spaces with your own AI for free. Keep projects in your browser, or ask your agent to export the project JSON and save it to your own files. Import it later to keep working. No Laydyne subscription or cloud project storage required.
Compare operational options through simulation, using the built-in AI or your own AI over MCP. Read and export saved results for free.
MCP / WebMCPExport / importThe agent saves and loads the file
Free
Your AI
Codex, Claude Code, or a compatible browser agent
“Build the airport terminal. Then compare three security lanes with five.”
create_spaceAirport terminal, 2 floors
edit_spaceSecurity lanes ×3
compare_process_options3 lanes vs 5 lanes
MCP / WebMCP
Laydyne Studio
3D model2D planSimulation
Compare3 lanes5 lanesIllustration: shorter is better
Run new simulation comparisons, test a range of conditions, and evaluate operational options, including through MCP. Reading and exporting saved results is free.
MCP: connect your usual agent
With a free account, your AI app connects with sign-in and works in an AI workspace saved on Laydyne, which you open in Studio. Agent links instead relay tool requests and results to an open Studio tab and store them only briefly for delivery.
WebMCP: work directly in the tab
A compatible browser agent calls the editing tools directly in the page, without the MCP server relay. Free editing tools need no account. Browser and agent support is required; WebMCP is experimental.
Local project storage does not mean offline use: loading the app, account and plan checks, and your AI provider can use the network. External AI data policies apply separately.
You can ask AI to create and change the space and workflow through conversation. Use the 2D/3D view to review what it made and check the assumptions. Manual editing is there for any adjustment you want to make yourself.
Can AI build a workflow beyond the examples?
Yes. AI can combine supported movement, arrivals, timed tasks, resource capacity, queues, and branches into a custom process scenario. This extends what you can model without a new template. It does not add new physics or execute arbitrary application code.
Can I try it before signing up?
Yes. Open Studio to explore sample spaces, edit layouts in 2D and 3D, and check routes for free. See Account for current features and usage limits.
Can I use Laydyne with ChatGPT, Claude or Codex?
Yes. After free sign-up, add one Laydyne URL to ChatGPT, Claude or another app that supports custom MCP servers, or connect Codex or Claude Code. Your AI inspects and edits the space and runs geometry checks in an AI workspace saved on Laydyne, which you can open in Studio; Studio does not need to stay open. Each AI app sets its own plan and region requirements. Modelling, route checks, and project JSON export and restore are free.
How accurate is the simulation?
Results come from a simplified, deterministic model: robots are circles moving at constant speed, and the same input always gives the same result. It does not model collisions or avoidance between robots or people (only capacities and queues), crowd physics, optimised lift group control (lift cars follow simple fixed rules), turning, or cleaning quality. Batteries, when a scenario gives them, drain in a straight line and charge at a constant rate. Dimensions read from photos and drawings are assumptions to check against measurements. It is for exploring options, not a calibrated prediction of a real site.
Does it calculate savings or recommend a purchase?
Studio can screen 1–8 robots against a cleaning coverage or transport completion goal. If you supply complete labor and robot costs, it compares the manual baseline, period costs, and break-even paid hours. It does not infer payroll savings or recommend a specific product; validate the model against the site before buying.
What happens to my data?
Your working space and conversation stay in this browser unless you export a project file or explicitly save a cloud project. See Account for cloud storage availability and limits. A cloud project includes the model, recent conversation, and images added as floor-plan underlays; saving is manual, not automatic sync. Account, session, and cloud project records use Cloudflare D1 (a local database during development). Built-in AI requests send the conversation, scene, and attached source to Laydyne’s AI service providers. When an AI app connects with sign-in, the project it edits is saved on Laydyne as an AI workspace until you delete it. Agent links relay tool requests and results through Laydyne servers, with temporary storage for delivery. Compatible WebMCP agents call editing tools directly in the page without that relay. Neither path is a promise of offline use. Attachments sent only for AI processing are not saved separately by Laydyne. Cloudflare delivers sign-in email.
Build your space with AI.
Connect your AI, describe your space, and ask for the first version.