Rebar detailing with AI in IFC models
With moddl you detail rebar right in the IFC model: bar layers, stirrups, welded mesh, trimming bars at openings and punching shear reinforcement. You name diameter, spacing and concrete cover in your own words, and moddl lays the bars into the concrete element, checks them and reports the steel quantity. It runs in the browser, with your own AI subscription.
01What you tell your AI about the rebar
You describe the reinforcement the way you would tell a colleague. Your AI turns that into calls to moddl. For example:
Reinforce the ground slab top and bottom with Q335A, concrete cover 5 cm at the bottom, 3 cm at the top.
Lower column: 8 Ø25 longitudinal, stirrups Ø10 / 15.
At the service opening, add 3 Ø16 at the bottom along each edge and a diagonal bar across every corner.
Which bars fall short of 3.5 cm cover?
Steel per element, in kg per m³ of concrete.
02How moddl lays rebar in the IFC model
Every laying belongs to one concrete element. With create_reinforcement your AI names the element, the kind of laying and its dimensions. moddl works out the bars and writes them into the model as IfcReinforcingBar or IfcReinforcingMesh. If the element moves, its reinforcement moves with it.
- Bar layers on one face of a slab, wall, beam or column, with diameter and spacing, also in part of the face or as a second layer behind the first.
- Welded mesh from stock, Q188A to Q636A and R188A to R524A, cut to fit the face, with longitudinal and transverse laps. Your AI can compare the waste in kg between variants before it writes.
- Stirrups along a beam or column, also in zones with their own spacing, as round links or a helix.
- Trimming bars at an opening, straight along both edges and diagonal across the corners, each running an anchorage length past the corner.
- Extra bars in the gaps of a layer and corner bars in a stirrup cage.
- Punching shear reinforcement from stud rails around rectangular columns. moddl measures the effective depth for them on the bending reinforcement over the column.
- Single bars with bending shape, hooks and mandrel, such as starter bars running on into a neighbouring element.
You give the concrete cover for the whole element or per face. moddl checks it while laying: if a bar falls short anywhere, moddl writes nothing and names the spot with the measured value.
moddl works out anchorage and lap lengths to EN 1992-1-1 or to DIN EN 1992-1-1/NA. If something is missing, such as the transverse lap of a mat, moddl takes the minimum of the code and flags it in the report as an assumption.
03Change diameter and spacing without redrawing
Every laying moddl made keeps its input in the model. With edit_reinforcement your AI names only what changes, say Ø14 instead of Ø12 or 15 instead of 20 cm spacing, and moddl lays it anew from there. The bar set keeps its GlobalId and its bar mark.
Whatever depends on that laying follows in the same revision: extra bars in its gaps, corner bars in its stirrups, the second layer behind it and stud rails whose effective depth was measured from it. If one of them no longer fits, it stays as it was and the report names it. Every change is a revision, and undo takes it back.
04How moddl checks concrete cover, laps and clashes
Before handover your AI checks the reinforcement with check_model. moddl measures on the bodies of steel and concrete:
rebar_cover: the concrete cover of every bar against your minimum, and every bar sticking out of the concrete.rebar_lapandmesh_lap: lap lengths against the code or your own value, such as 50 d, and mats that overlap too little.rebar_clash: bars running through each other, also between layers, stirrups and stud rails.rebar_spacing: the clear spacing per layer to EN 1992-1-1 8.2, with aggregate size if you give it, and the largest axis spacing against your value.- It also finds sets laid twice, mandrels that are too small, faces without reinforcement and too many mats stacked on one spot. If you ask, moddl checks whether a bar is too long or too heavy to transport.
These checks also run on reinforcement exported from another program. moddl then proposes from the geometry which element a bar belongs to. Where the reinforcement meets pipes or other elements, clash detection finds it.
05Read off steel quantity and reinforcement ratio
For each concrete element your AI reads everything in one call with find_elements and the field ReinforcementSummary: the layers per face with diameter, spacing and provided steel area in cm²/m, the least concrete cover, laps, clashes, bar marks, steel mass and the reinforcement ratio in kg per m³ of concrete. Your AI compares the provided area with the required one from your structural design.
compute_quantities gives count, length and mass of the bars, plus anchorage and lap length per set and the waste of a mat laying. How quantities come about in moddl is covered under quantity takeoff.
For every number you see where it comes from. Steel mass, for example, uses a density of 7,850 kg/m³, and the answer flags that as an assumption.
06What moddl does not do for reinforcement
- moddl does no structural design. The required reinforcement comes from your calculations. moddl lays what you name and measures the steel area provided.
edit_reinforcementonly changes layings moddl made itself. Reinforcement from other programs it reads and checks.- Reinforcement drawings and bar bending schedules are in the works. Until then your AI reads bar marks, bending shapes and masses from the model as a list, and you download the IFC file with the reinforcement.
07How to start
Frequently asked questions
What are the requirements?
You need an AI assistant with MCP support, for example Claude or ChatGPT. ChatGPT can also be used on the free plan. You connect it with your own account, without an API key: Edit BIM with your own AI subscription.
How do I connect Claude or ChatGPT?
moddl walks you through the connection step by step in the viewer. You can also read the guides here: Connect Claude to moddl and Connect ChatGPT to moddl.
What does it cost?
During the beta phase, expected to run until June 30, 2027, moddl is free of charge. We will usually announce the end of the beta at least two weeks in advance. Your AI assistant is subject to your provider's terms.
Does the AI get access to my IFC file?
No. The AI receives only the answers to its requests and reports on the changes it has made. These answers may contain content from the model, such as element names and properties.
Can changes be undone?
Yes. Every change creates a new version, and every earlier version can be restored. The uploaded original file stays unchanged.
Can I use the result in my BIM software?
Yes. You can export the model as an IFC file at any time. moddl writes only the parts that changed; all other values are preserved unchanged.