Guide

AI for Navisworks: Clash Detection & Model Review with Claude

"AI for Navisworks" now means something concrete: you can drive Navisworks with an AI assistant in plain English — create and run clash tests, summarise the results, work with TimeLiner 4D schedules, search the federated model and export reports — without clicking through dialogs. This guide shows what that looks like in practice, with real prompts.

What "AI for Navisworks" means today

Model coordination AI comes in two flavours. One talks about Navisworks — it can explain what a clash test is, but you still set it up yourself. The other operates Navisworks: creates the test, runs it, groups the results and writes the report. The second is the one worth having.

HuskyBIM takes that second approach. It connects Claude Desktop to Navisworks through the Model Context Protocol (MCP) — an open standard that exposes Navisworks' capabilities to the AI as callable tools. You describe a task; Claude picks the tools and the connector runs them on your open model. HuskyBIM for Navisworks ships 82 such tools.

What you can genuinely automate

Here are real prompts, grouped by the kind of work they cover.

Clash detection
Create a clash test between the Ductwork and Structure selection sets with a 25 mm tolerance, run it, and summarise the results.
Group the clashes in "MEP vs Structure" by level, mark the ones inside shafts as reviewed, and export a clash report.
TimeLiner and 4D
List the TimeLiner tasks with their planned dates, and flag any task whose linked items are empty.
Create a TimeLiner task for the Level 3 concrete pour linked to this selection set, then export the schedule to CSV.
Search and review
Find every item whose "Status" property is "Temporary", highlight them in red, and save a viewpoint.
Isolate everything on Level 3 and set a section box around it so I can review the MEP routing.
Model intelligence and export
Give me a model summary: file links, item counts by category, and the model units.
Export the current view as an image, and dump the properties of the selected items to CSV.
Point clouds
List the point clouds in the model, report their bounds, and measure the scan-to-model distance for this wall.

The recurring patterns are clash workflows (set up, run, triage, report), 4D schedule reads and edits (TimeLiner tasks and dates), model QA (property searches across the federated tree), and reporting (images, CSV dumps, IFC/NWD export). The clicking-heavy coordination work is exactly what AI handles well.

It runs the clash tests, not just reads them. The connector creates clash tests, executes them, reads and summarises results, and can set clash-result statuses — so a full triage round-trip happens in one conversation.

How it works

The chain is short: Claude Desktop hosts the HuskyBIM extension where you type; the connector — a small local service — receives Claude's tool calls; and it drives Navisworks through its .NET API on your open federated model. Everything runs on your Windows machine; Claude Desktop sends your request text and the tool results to Anthropic's models to reason over them, the same as any Claude chat.

Honest limits

Try AI for Navisworks — free

Create a free HuskyBIM account, install the connector for your Navisworks version, open your federated model, and start giving Claude instructions in plain language.

Frequently asked questions

Can AI for Navisworks actually run clash tests?

Yes. HuskyBIM for Navisworks creates clash tests, executes them, reads and summarises the results, updates clash-result statuses and exports reports. Clash features require Navisworks Manage, which is where clash detective lives.

Which Navisworks versions are supported?

Navisworks 2026 and 2027 on Windows, as separate installers — install the one matching your version. Both connect Navisworks to Claude Desktop through the Model Context Protocol.

Does it work with TimeLiner and 4D schedules?

Yes. It reads TimeLiner tasks, dates and simulation settings, creates and updates tasks, links them to selection sets and exports the schedule to CSV — so schedule QA and 4D housekeeping can happen in plain language.

Is my model data safe? Where does it run?

The connector runs on your own Windows machine and talks to Navisworks over a local connection. To answer your prompts, Claude Desktop sends your request text and the data the tools return to Anthropic's Claude models, the same as any other Claude conversation.