NexusDBP takes a building permit from the first question to the issued permit. An AI assistant guides applicants through every step, and a deterministic compliance engine checks BIM models, 2D drawings and the site against the building code. Same data, same result, every time.
From the first question to the issued permit: the assistant, the six steps, the checks, the BIM viewer, issues, the site, 2D drawings, the rule builders and the integrations.
Applicants, reviewers and the people who maintain the building code work in one place. Every result points to the code section it comes from.
Ask a question, drop your files, or let it guide the whole application. It explains failures in plain words and starts uploads and checks for you.
Six steps from project info to submission: models and drawings, data quality, site and spatial checks, building checks, summary and submit.
Every model is validated against the IDS of its code set first. Missing data shows up as missing, never as a false failure.
Every rule and calculation of the code set runs on the model. Each run records the engine version, the models and the code set, so it can be repeated.
Failed elements light up in 3D, or on floor plans cut from the model itself, level by level, side by side with the 3D view.
Raise an issue on the model with its elements, view, markup and snapshot. Reopen it later and the same view comes back.
Setbacks, height, coverage, floor area ratio and utility connections, checked against live GIS services on a 3D map of the parcel.
PDF, DWG or DXF sheets are classified, scaled and read: dimensions, tags, schedules and sections. An unclear reading goes to review, never to fail.
ETABS and SAFE packages are read and placed on the model, with our own plausibility checks and estimates of the results.
Describe a rule in plain words and AI drafts it, or build it yourself: visual, diagram, SQL, spatial and 2D drawing rules. A person approves every rule.
Building codes by chapter and section, checks with versions, and code sets per permit type. Every check is pinned to a section of a code.
Justify a failure where one is still needed, see the required documents, preview the compliance report and submit. The files and results are locked once submitted.
Each rule always returns the same result when faced with the same data: no inconsistencies and no hallucinations.
Guides, explains, drafts
Decides every pass and fail
AI never decides a pass or a fail. The parcel and the zoning come from GIS, not from AI.
The assistant is the front door: ask a question, drop your files, or let it guide the whole application. It fills in the application as you answer, then creates the project and takes you to step one.
It says what to do next and starts uploads and checks for you. Prefer the buttons? They are all there.
The parcel, the zoning and the district are read from the authority's GIS services, not guessed.
It knows which documents each permit type needs, straight from the authority's own configuration.
Frontier APIs or self-hosted models, routed per task. Data that must stay local can be kept on local models.
Every rule and calculation of the code set runs on the model. Same model, same code set, same result, every time. The assistant explains the failures in plain words; the engine decides them.
Models are validated against the IDS of the code set before any check runs. Failing elements export as BCF for the authoring tool.
Each run records the engine version, the model versions and the code set, with its history kept per project.
Every pass and fail points to a code section, with the elements it applies to, one click from the BIM viewer.
Fix the model and re-upload; justify a failure only when it is the exception. The reviewer decides on every justification.
Failed elements light up in 3D, straight from the check results. Or on floor plans cut from the model itself, level by level, side by side with the 3D view.
Pick a result and the plan moves to the level that holds it, while the 3D view frames the same elements.
Measurements and notes on the plan, a section box in 3D, and a model browser with every property.
ETABS and SAFE packages are read and placed on the model, coloured by utilisation, with our own plausibility checks and estimates of the results.
Take the results back to the authoring tool as a BCF file.
Frame the view in 3D, 2D or Split, measure and add notes. The snapshot is taken for you. Open the issue later and the same view comes back, down to the measurement on the plan.
Each issue keeps its elements, its view, its markup and the snapshot.
Type, priority, status, assignee, due date, comments and attachments.
Everyone on the project sees the same issues, each one linked to its elements on the model.
Setbacks, height, coverage and utility connections, checked against live GIS services, on a 3D map with the parcel, the zoning and the networks around it.
The model sits on its parcel, with the neighbouring buildings and terrain around it.
Each failed check shows its measurement on the site: the setback, the height, the distance to the manhole.
The GIS data is read at check time and stored with the run, so the result can be reproduced later.
Many engineers out there are still using 2D drawings. You can check the building code compliance even on top of your 2D drawings.
Sheets are classified and scaled from the title block: plans, sections, elevations, details and schedules.
Dimensions, tags and schedules are read, and the checks run on what is read: a door width on the plan, a stair riser on the section.
An unclear reading goes to review, never to fail. Results are marked on the sheet itself.
Describe the rule in plain words and AI drafts it, as a visual rule and as SQL. A person approves; the engine runs it without AI. Or build it yourself.
From the requirement to a draft rule, checked against a real model before anyone publishes it.
Element sets, filters and conditions, as a form or as nodes in a diagram. Nodes in, SQL out.
Calculations in SQL, rules against GIS layers, and rules that read 2D drawing sheets.
Every check is pinned to a section of a building code, published in versions, and grouped into code sets.
For the authority, a review that starts from facts. For the applicant, no surprises at submission.
Click a screen to see it full size.
Each role sees exactly what it needs, inside the authority's own organisation.
Creates projects, uploads files, runs checks, justifies failures and submits
Reviews submissions, accepts or rejects justifications, logs issues and issues decisions
Maintains building codes, checks, IDS and code sets, and publishes check versions
Users, permit types, GIS services, AI models, MCP and deployment settings
AI agents work through the MCP server with the same permissions as the person who connected them.
From a single house to a public building, from a 2D drawing set to BIM models per discipline.
Receive complete applications, review from the engine's results instead of from scratch, and issue decisions that point to the code.
Check the design against the code before you submit, fix what fails in your own tool, and justify only the exceptions.
Submit the ETABS or SAFE package with the model, see the results placed on it, and the plausibility checks next to them.
Turn the building code into versioned, testable checks, with AI drafting and people approving, and publish them per permit type.
Your AI models, your agents, your GIS services and your sign-in: NexusDBP connects to what the authority already runs.
Local or frontier, routed per task under your data policy. AI never decides a pass or a fail.
Your agents run checks and read results with your permissions. A person confirms every submission in NexusDBP.
Parcels, zoning, roads, utility networks and height limits, read live at check time and kept with the run.
People sign in with the authority's directory, and its groups map to NexusDBP roles.
Models, data requirements and issues in buildingSMART standards; drawings as they are produced today.
Read from the engineer's own analysis files and placed on the model, with plausibility checks next to them.
Run NexusDBP as a service, or install it in the authority's own data centre. Only who runs it changes.
Hosted and run by INS, with the authority as its own organisation and its own sign-in.
Installed with Docker Compose on the authority's servers, with an offline licence: no internet access needed, no call home.
Self-hosted language models on the authority's own GPU servers, so no document has to leave the building.
Book a walkthrough with your own code set, models and drawings, and see how NexusDBP takes a permit from the first question to the issued permit.