BetaThe ClashFree public beta is open. Free 15-day trial, Revit 2023–2027.Get your invite →
  1. Home
  2. FAQ

33 answers

Frequently asked questions.

The short answers. For every screen of the add-in, see the help center; for install and activation problems, support.

Product

What is ClashFree?

ClashFree is an AI-native clash resolution add-in for Autodesk Revit. You import a Navisworks XML or BCF clash report, choose your trade and send clashes to AI. For each clash ClashFree lists every feasible fix, scores it, applies the best one in your model, verifies it against the host and linked models, and waits for your Keep or Reject.

Who is it for?

BIM coordinators, MEP designers and modellers, BIM managers, MEP engineers and contractors who resolve clash reports in Revit, for HVAC, plumbing and drainage, electrical and fire protection.

Does ClashFree detect clashes?

No. Keep running your clash tests in Navisworks or your BCF tool and import the report; several files merge into one list. Without a report you can pick the two clashing elements in Revit. ClashFree takes it from there to a fixed, verified model.

Does it replace Navisworks?

No. Navisworks, or your BCF tool, finds the clashes; ClashFree resolves them in Revit. It reads the report, sets aside the noise (duplicates, approved, within tolerance, an element against its own run), groups the rest by root cause, and fixes your trade's runs in the model. Keep your clash tests where they are.

Which elements can it move?

Pipes, ducts, conduit and cable tray of your trade in the host model, with their fittings. Anything else, and anything in a linked model or of another trade, is a read-only obstacle; when the right fix is theirs, ClashFree flags it for a person with the reason.

What kinds of fixes can it make?

Vertical and sideways offsets with 45° or 90° fittings and moves of a short run with open ends. It can also drag a run back with its end fittings (the neighbours stretch), slide or tilt a tee, shift a run that ends in a dead end, reroute around obstacles, and make area fixes that clear several neighbouring clashes together. The pane's AI and Claude over MCP have the same fixes; each one is listed, scored and verified, and when none verifies the clash is flagged for a person with the reason.

Is clash resolution fully automatic?

No, by design. ClashFree automates the routine of each fix: choosing it, splitting, offsetting, fitting, reconnecting and re-checking. Every fix still waits for your Keep or Reject, and Keep is always your click.

What are To fix, Coordination, Ignored and Cleared?

To fix holds your trade's runs that ClashFree can move. Coordination holds clashes nothing of yours may move: another trade or a person decides. Ignored holds duplicates, clashes approved or resolved in the report, within tolerance, or set aside by you. Cleared holds clashes that no longer clash, because of your fix, another kept fix, or Revit finding the elements apart.

AI and judgement

Does the AI draw the geometry?

No. The AI never outputs coordinates. It picks one of the scored options, or a strategy with millimetres, an angle and a direction, and explains why. Deterministic code builds the geometry and Revit verifies it before you see it.

If the AI is just Claude, what stops it from wrecking my model?

The plugin, not the AI. The AI only chooses: an option from a list ClashFree scored, or a strategy with millimetres and a direction. Everything that decides what may happen is deterministic code in the add-in: which runs are yours to move (linked models and other trades are read-only), the hard limits no score or AI can break (ceiling lines, structural walls and beams, slabs, gravity falls), the scoring, building the geometry, and verifying it against the host and linked models. Every fix runs inside one Revit transaction group as a preview: it is verified before you see it, and Reject restores the model exactly. Nothing is committed until you click Keep. Swap the AI for a weaker one and the worst case is a fix that fails verification, never a broken model.

Is this just ChatGPT with a Revit prompt?

No. A chat model on its own cannot see your model, cannot edit it safely and cannot check its own work. ClashFree reads both runs through their fittings, the free space, walls, slabs, neighbouring runs and your coordination profile, lists every feasible fix and scores it, and the AI picks among those with its reasons. The add-in then builds the fix, verifies it in Revit and rolls it back if anything new clashes. The AI is the judgement in the middle; the reading, the rules, the edit and the check are ClashFree's.

How does ClashFree decide which service moves?

By the cost of the change. Every run has a system class (supply duct, smoke extract, sprinkler, gravity drainage, hydronic, domestic water, condensate, cable tray, conduit and more) with change costs scaled by size. The cheaper change gives way: a 50 mm water pipe moves before an 800 mm supply duct, whose reroute would need an airflow re-check.

How are fixes scored?

Each option is priced by the fittings it adds, the length it adds, how far the run moves, the system class's factors, the walls it would newly cross, leaving the trade's height zone and new low or high points; matching how neighbouring runs already pass the obstacle earns a bonus. Hard limits remove options before any score counts. Lower is better.

What are hard limits?

Rules no score can outweigh and no AI override can break: a project's ceiling line, structural walls and beams, slabs at risers, linked models and other trades, gravity drainage and the fall of a sloped line. An option that breaks one is removed and listed with the reason, so you can see why it was not chosen.

Can I teach it my project's rules?

Yes. The coordination profile has three layers: base (general practice), project (ceiling heights, height zones per trade, clearances, wall classes, system classes) and you (soft weights only). Import a project profile in the pane; when you keep the second-ranked option for the same reason three times, ClashFree suggests a preference, and never applies one silently.

Does it learn how we work?

Yes, in the open. Your grouping decisions (clashes you set aside, merged or moved) carry over to the next import of the same report set. Neighbouring runs that already pass an obstacle become precedents that a fix at the same level is scored up for. When you keep the second-ranked option for the same reason three times, ClashFree suggests a preference for your profile; it never changes a weight silently, and every fix page shows which terms decided it.

What happens when a fix fails verification?

The reason goes back to the AI and the next-best option is tried, up to three, before the clash is flagged for a person with the reasons and a limit code such as dead end, slab or crossed wall.

Does it tell me what an engineer must re-check?

Yes. Every fix carries a bill: fittings added and removed, length, parts moved, and what it impacts, such as an airflow re-check on a duct, the hydraulic calculation on a sprinkler pipe, a vent or drain for a new high or low point, or a firestop on a rated wall. ClashFree simulates nothing and never claims a system is fine.

Which AI model does it use?

ClashFree is provider-neutral; the beta uses Anthropic's Claude on your own API key, and the pane shows which model and the expected cost before every run. Over MCP, Claude Desktop or Claude Code does the reasoning on your Claude plan.

Data and privacy

Does my model leave my computer?

No. Your Revit model, clash reports and project, file, user and machine names are never sent. The AI gets a text description of the clash and its surroundings, straight from your PC to your AI provider.

What does the beta send to ClashFree?

Usage analytics (counts and codes, a condition of the beta), scrubbed diagnostics once a day, and, unless you turn it off in Settings, scrubbed Keep and Reject records of the clash context and the proposal. Names, paths and identifiers are hashed or removed first. Details are in the beta data terms.

Where is the data stored and for how long?

On Cloudflare in the EU. Usage events and shared clash content are kept 13 months, diagnostics 90 days. Settings, Beta, Delete my data on the server removes this install's data at any time.

Where is my API key stored?

Encrypted for your Windows user on your PC. It goes only to your AI provider; ClashFree never sees it.

Do you sell my data?

No. We use it only to run the beta and improve ClashFree. See the privacy policy.

Trial and pricing

How much does the beta cost?

Nothing. The beta is a free 15-day trial per e-mail address with every feature. AI usage is billed by Anthropic to your own key, or runs on your own Claude plan over MCP.

When does the trial start?

At your first activation in Revit, not at sign-up. Reinstalling does not restart it. Days are counted in UTC.

What happens when the trial ends?

The add-in locks. A fix preview that is still open can still be rejected, so your model is never left half-changed. Nothing is charged.

Can I extend my trial?

If you need more time on a real project, contact us and tell us what you are testing; we extend beta trials case by case.

What will it cost after the beta?

Plans and pricing are coming soon. Until then, a licence for you or your team is arranged directly: contact us and we invoice; companies pay by bank transfer.

Install and Revit

Which Revit versions are supported?

Revit 2023, 2024, 2025, 2026 and 2027 on Windows 10 or 11. One installer finds the versions on your PC and you tick the ones that get ClashFree.

Do I need administrator rights?

No. ClashFree installs for your Windows user only. Close Revit first.

How many machines can I use?

Your invite code activates up to 3 machines with the same e-mail address.

Is there a Mac version?

No. Revit runs on Windows, and so does ClashFree.

Can I undo a kept fix?

Yes. Keep commits the fix as one Revit undo step, so Revit's Undo reverts it like any other change.

Claude and MCP

Does ClashFree work with Claude?

Yes. ClashFree ships an MCP server, and the installer can connect it to Claude Desktop or Claude Code. Claude can then list groups, open clashes, read both runs, the free space and the scored options, and propose fixes through the same checks.

Can Claude change my model without me?

No. Claude's fixes are verified previews, exactly like the pane's. Keep stays your click in the pane, and each clash has a budget so Claude cannot loop.

Do I need an API key to use Claude over MCP?

No. Over MCP the reasoning runs on your Claude plan in Claude Desktop or Claude Code; the API key is only for the pane's own AI and for running the AI queue from the pane. Send to AI needs no key, and Claude can take the queued items itself.

Another question?

A person who builds ClashFree answers within two working days.