Electrical / runnable MCP tool

Run the Segmented 90 Calculator in your AI assistant

Through the free Intry MCP server, Claude, ChatGPT, or Cursor runs this exact calculator and hands back a verdict and the cited source, not a guess. Tool name: segmented_bend.

How it computes the answer

Stated by geometry, because no bender document this site holds prints a segmented-bend table: developed length of the sweep = target radius × π / 2; mark spacing = developed length / shots; the first mark sits half a space past the start of the bend, and every shot is bent 90 / shots degrees at its mark. Each shot's arc must fit on its own segment at the NEC Chapter 9 Table 2 bend radius; a refusal names the smallest target radius that works at this shot count and the fewest shots that work at this radius.

What it cites

Your assistant gets the same verdict and the same citation a technician gets on the interactive calculator, because both call one engine. These are the sources that engine reads.

Bender-manufacturer offset multiplier table

Distance between the two offset marks = offset depth x the multiplier: 6.0 at 10 degrees, 2.6 at 22.5, 2.0 at 30, 1.4 at 45, 1.2 at 60. Printed as a constant beside the shrink column, under the heading 'applies to any size conduit', in the Greenlee Offset Table (Hand Bender Instruction Manual IM 1071 REV 4 5/98), in the Greenlee SITE-RITE Hand Benders manual, and in the Gardner Bender 960-Series How to Bend guide; all three agree cell for cell on the values (Gardner Bender's angle column misprints 22-1/2 as 21-1/2; its own instructions on the same page say 22-1/2), and Klein's Zip Guide distance-between-bends table is computed with the same constants (8 inches of depth at 60 degrees prints 9-5/8 inches, which is 8 x 1.2 to the nearest eighth). Each constant is 1/sin(angle) taken to a tape-measure value: four of the five are the cosecant to one decimal (2.61, 2.00, 1.41 and 1.15 unrounded) and 10 degrees is rounded up from 5.76 to 6.0, and the calculator applies the chart value at a charted angle so its marks match the table cast on the bender. At any other angle it applies the exact 1/sin and says so.

Bend geometry (trigonometry)

Where no chart row exists the engine works from geometry: the offset multiplier is 1/sin(angle), the three-point saddle center-to-outer distance is depth/sin(half the center angle) at a center no held saddle table prints, a kick's travel is kick height/sin(kick angle), and a rolling offset is the Pythagorean combination of the rise and the roll. No held document prints a kick constant, so the kick is geometry alone: the kick mark sits kick height/sin(kick angle) past the 90 mark, and that mark distance is the bend the engine checks and the 3D draws. The receipt names which basis produced each answer.

Bender-manufacturer 3-point saddle table

Distance from the center mark to each outer mark = saddle depth x 2.5 for a 45 degree center (22.5 degree outer bends), printed as a saddle table in four held documents that agree cell for cell: the Greenlee Hand Bender Instruction Manual IM 1071 REV 4 5/98 ('Table for 45 degree Saddles', 1 in -> 2-1/2 in, 2 in -> 5 in, 3 in -> 7-1/2 in, 4 in -> 10 in, 5 in -> 12-1/2 in, 6 in -> 15 in, 'for each additional inch add 2-1/2 in', under 'applies to any size conduit'), the Greenlee SITE-RITE Hand Benders manual (the same table), the Gardner Bender 960-Series How to Bend guide ('outer bend marks at a distance of 2-1/2 in x the object size') and the Klein Tools Conduit Bender and Angle Setter Guide (MKT200815 Rev. 09/20 A) Zip Guide for Saddles (the 45 degree column). Klein's 60 degree column (30 degree outer bends) prints 2 in per inch of depth: 2, 4, 6, 8, 10, 12. The calculator applies the table value at those two center angles so its marks match the table on the bender; exact 1/sin of the outer angle is 2.61 at a 45 degree center (the table is that value rounded for a tape measure, and the marks at the table spacing lay out a sharp-corner peak 4.3 percent below the requested depth, before the bend radius rounds the top) and exactly 2.00 at 60, and both are shown beside the applied value. The peak the calculator and its 3D report is the centerline height once the center arc is laid at the NEC Chapter 9, Table 2 radius, which sits R x (sec(outer angle) minus 1) below that sharp corner; pure geometry, no sourced constant. One printed cell is refused rather than matched: Klein's 1 inch row at a 60 degree center (2 inches from center) cannot be bent in any EMT size this calculator carries, because the 30 degree outer arc and the 60 degree center arc on the Table 2 radius overlap before the marks are 2 inches apart, and the calculator says so instead of printing it. No held document prints a saddle row for a 30 or 22.5 degree center, so those saddles apply the exact 1/sin and say so. The saddle shrink is the offset chart's shrink at the outer angle, which is also what the same tables print in their shrink column.

Bender-manufacturer shrink-per-inch table

Shrink per inch of offset: 1/16 in at 10 degrees, 3/16 in at 22.5, 1/4 in at 30, 3/8 in at 45, 1/2 in at 60. Fetched from three bender-manufacturer guides that agree cell for cell: the Greenlee Offset Table (Hand Bender Instruction Manual IM 1071 REV 4 5/98), the Gardner Bender 960-Series How to Bend guide, and the Klein Tools Conduit Bender and Angle Setter Guide (MKT200815 Rev. 09/20 A). Greenlee and Klein additionally print the same absolute shrink-per-depth table as each other, cell for cell, including a shared rounding anomaly at 22.5 degrees and 9 inches of depth where both print 1-3/4 inch against a pattern value of 1-11/16. Two documents carrying one anomaly is a common convention, not independent confirmation, so treat the three guides as corroborating the trade standard rather than as three separate derivations of it. The 15 degree row is the one value none of the three publishes, and it is marked as unsourced rather than attributed. This is the field table. It does not equal tan(half the bend angle), and that difference is published beside it on the shrink chart rather than smoothed over: tan(half-angle) is the shrink of a bend modelled as a sharp corner with no radius, so it runs above the chart, while a real bender lays an arc and shrinks slightly less than the sharp-corner figure. We apply the manufacturer chart wherever one exists, and fall back to the sharp-corner geometry only for angles no manufacturer publishes, which is what lets the calculator answer any angle.

Hand-bender take-up (EMT) and gain (every type)

Take-up is the distance from your mark to the start of the bend. Two held guides print the same three EMT values and both match this engine exactly, 1/2 in at 5 in, 3/4 in at 6 in and 1 in at 8 in: the Klein Tools Conduit Bender and Angle Setter Guide (MKT200815 Rev. 09/20 A) in its Bender Take Up Table, and the Gardner Bender 960-Series How to Bend Guide in its deduct chart. Both of those tables stop at 1 in, so the 1-1/4 in take-up is covered by neither, is carried as a field value, and is marked as unconfirmed everywhere it is printed. Greenlee is deliberately not cited for take-up: both Greenlee manuals we hold publish no take-up or deduct value at all and instruct the electrician to read the deduct off the bender casting, which is why the calculator also accepts a deduct you enter from your own tool. A deduct you enter replaces the take-up and nothing else, because it does not tell us your shoe's radius, and the answer says which of the two produced it. The field gain the engine subtracts from leg lengths is derived from the radius and the outside diameter, never hand-typed, so the two gain references cannot be mixed up.

These are geometry and published field-chart values, computed and applied by the engine, not a laboratory measurement. Bender take-up varies by tool and shoe, so field-verify the first bend of a new setup against the actual conduit. The bender in your hand has the final say.

The full receipt, with every source and how each is cross-checked, is on the interactive Segmented 90 Calculator.

Ask your assistant

The one rule that makes it reliable: name Intry in your question. That is what tells Claude to run the segmented_bend tool and answer with the real number, the verdict, and the citation. Leave Intry out and Claude may just guess from memory, which is exactly what you are trying to avoid. Type something like:

Inputs

Add Intry to Claude

One paste and Claude can run this tool and every other Intry calculator, right from your phone.

Using ChatGPT, Cursor, or a dev tool? See the developer setup on the main page. The endpoint is https://www.intrysys.com/api/mcp.

Every answer comes with the bender chart or geometry it was computed from and the NEC Chapter 9 Table 2 radius, so you can check it or show an inspector. The numbers come from the same calculator we run on intrysys.com, and we re-check every one automatically before anything goes live, so the number Claude hands you matches the code.

Also on Intry