BlenderBlender and AI

Letting AI build a modifier stack: Array, Mirror, and Bevel over MCP, then fixing it by hand

Through the official Blender MCP, ask AI to stack Array, Mirror, and Bevel modifiers on one fence picket, read the real result in the Modifiers tab, and fix the picket count the AI read differently with one number. Two common mistakes, unapplied scale and the Mirror origin, are covered too.

AZTechnology14 min read

한국어English

Based on: Blender 5.2 LTS (screenshots from 5.2.1, English interface, default shortcuts), the Blender Lab official MCP server and add-on 1.0.3, and Claude Code as the AI app. Checked in November 2026. All images are real Blender screens. Red numbers and boxes were added for explanation.

If you hand modifiers to AI, the stack is built in under a minute. But the AI may read your words differently from what you meant. In this guide, you ask AI to stack Array, Mirror, and Bevel modifiers on one fence picket through the official MCP, read the real result in the Modifiers tab, and fix it with one number.

A modifier is like a filter in a photo app. The original photo stays as it is, and the effect is shown on top. Filters stack and apply in order, and so do modifiers, from top to bottom. This list is called the stack. If modifiers are new to you, read What is a modifier? first.

When you finish this guide, you will be able to do the following.

  1. Write a question that asks AI to build a modifier stack.
  2. Read the stack in the Modifiers tab and count the result.
  3. Find in the answer where the AI read your question differently, and fix it with a number.
  4. Tell apart common mistakes such as unapplied scale and the Mirror origin.
  5. Change the shape with one number and see why an unapplied stack is useful.

Before you start

You need: an AI app connected to the official MCP (for example, Claude Code), and Blender 5.1 or later with the server running. If you have not connected yet, see Installing the official Blender MCP and Connecting the official Blender MCP to Claude Code. If you have not picked an AI app, Choosing an AI tool for Blender MCP compares them.

  1. In Blender, open a new file with File > New > General. It has one default cube.
  2. In the Properties editor on the right, open the Modifiers tab (wrench icon). The modifiers the AI adds show up here right away.
  3. If this is a file you care about, save it first. Changes the AI makes with code do not always undo in one Ctrl+Z. For more checks, see Safety checks before handing Blender to AI.

1. The question sent to the AI

Starting point: the default cube in a new file (File > New > General).

Do this: send the following to your AI app as it is.

Turn the default cube into one fence picket, use Array to make 6 to the right, use Mirror to make the same on the left, and use Bevel to round the corners a little. Do not apply the modifiers; just stack them.

The last sentence, "do not apply, just stack," matters. Applying is like baking the filter into the photo. Once baked, you can no longer change the count or the roundness with a number. A stack lets you keep editing the AI's result in the panels.

The AI does this work with the code tool of the official MCP. When this question was sent to Claude Code, the AI did the following. It took about 50 seconds, and at the end the AI took a screenshot of the viewport to check the shape.

StepWhat the AI didValues
1Renamed the cube to Fence_Picket
2Moved the cube's points to make a picket shapewidth 0.1 m, depth 0.04 m, height 1.2 m, bottom at height 0
3Shifted the picket to the right of the origin0.125 m (the object origin stays at the world origin)
4Added ArrayCount 7, Constant Offset X 0.25 m
5Added MirrorAxis X, merge off
6Added BevelAmount 0.008 m, Segments 3, Limit Method Angle

Front view of the AI's first result. Fourteen fence pickets stand along the X axis, and the Array panel in the Modifiers tab on the right shows Count 7 and a checked Constant Offset. Box 1 on the Count row, box 2 on the Constant Offset row Figure 1. The AI's first stack: 1) Array Count 7, 2) Constant Offset turned on (Blender screenshot)

2. Reading the result: the count and the last lines of the answer

Starting point: continue from the Fence_Picket the AI made in section 1.

Do this:

  1. Click Fence_Picket in the Outliner to select it.
  2. In the Modifiers tab, check that the panels run Array, Mirror, Bevel from top to bottom. The icon in each panel header tells them apart.
  3. Read Count in the Array panel. In Figure 1 it is 7.
  4. Count the pickets in the viewport. Figure 1 has 14.

The question said "6 to the right, the same on the left," so if you expected 12, there are two extra. The reason was in the last lines of the AI's answer. The AI read "6 to the right" as the original picket plus 6 more and set Count to 7. It added that if you wanted only 6, you can change Count to 6.

Read the answer to the end: AI often settles an unclear word one way, finishes the job, and explains that choice in the last paragraph. Do not stop at the screen. Read paragraphs such as "things to check."

Also look at the offset next to Count. The AI turned off Relative Offset (a multiple of the object width) and turned on Constant Offset (a measured distance). Click the arrow left of Constant Offset to open it, and Distance X reads 0.25 m. The picket is 0.1 m wide, so there is a 0.15 m gap between pickets.

Do this: check the scale and the origin too.

  1. Open the Object tab (orange square icon).
  2. Read Location X, Y, Z and Scale X, Y, Z.

The Object tab. Location is 0 m on all axes (box 1) and Scale is 1.000 on all axes (box 2). In the viewport, a small orange dot for the object origin sits on the floor between the middle pickets (box 3) Figure 2. The Object tab of the AI's result: 1) Location 0, 2) Scale 1.000, 3) the object origin at the world origin (Blender screenshot)

Scale is 1.000 on all axes and the origin is at the world origin. The AI did not change the object's scale values; it moved the points directly to make the picket. Because of these two things, the two common mistakes in section 4 do not appear in this result.

3. The fix: twelve pickets with one Count value

Starting point: continue from the Fence_Picket in section 2.

Do this:

  1. Open the Modifiers tab.
  2. Double-click the Count field in the Array panel.
  3. Type 6 and press Enter.

After changing Count to 6. Twelve fence pickets stand six on each side of the origin at even spacing, with box 1 on the Count field. The status bar at the bottom shows Verts 1,152 Figure 3. Count 6. Twelve pickets stand six on each side of the origin (Blender screenshot)

The pickets change to six on each side, twelve in all, right away. Verts (point count) in the status bar at the bottom drops from 1,344 to 1,152. One picket has 96 points, so 1,152 is twelve pickets. You do not need to ask the AI again or rerun any code. That is thanks to "do not apply" in section 1.

4. Two common mistakes: scale and the Mirror origin

Starting point: continue from the Fence_Picket in section 3. In this section you only read and check.

In this run, the AI avoided the two mistakes below. But they are common with other questions or other AI apps, and when they happen, the cause is hard to find from the screen alone. Knowing them in advance tells you what to look for when you read a result.

Mistake 1: the bevel is rounded on one side only because scale is not applied

If you make the picket by changing only the object's Scale values (for example X 0.1, Y 0.05, Z 1) instead of moving the points, modifiers work on the original cube (a 2 m box) before the scale is multiplied in. The corner cut on the original cube is then squashed to 0.1 times in X and 0.05 times in Y on screen. It is like stamping with a rubber stamp and then pressing the stamp flat from the side.

Two close-ups of the top of a picket. Panel 1, before applying scale, has thin bands only in the vertical direction and sharp top corners. Panel 2, after applying scale, has all four top corners evenly rounded Figure 4. For comparison, the same bevel (Amount 0.02 m) on a picket flattened with Scale values: 1) before applying scale, only the vertical direction is cut, 2) after applying scale, all four corners are evenly rounded (Blender screenshot)

How to spot it: suspect this when Scale in the Object tab is not 1.000. To fix it, place the mouse over the 3D Viewport, press Ctrl+A, and choose Scale. The current size is baked into the point positions and Scale goes back to 1.

Mistake 2: the middle pickets overlap because the Mirror origin is in the middle of a picket

The Mirror modifier flips around the object origin. If the origin is in the middle of the first picket, the mirror line runs through the middle of that picket, and the picket and its mirror image sit in the same place. It is like a person standing right against a mirror and overlapping their reflection. Then even if twelve pickets are made, you see eleven.

How to spot it: if one picket seems to be missing but the point count in the status bar is for twelve, they overlap. In this result the origin is at the world origin and the picket is shifted 0.125 m, so nothing overlaps (box 3 in Figure 2). If they do overlap, move the origin to the world origin (Object > Set Origin > Origin to 3D Cursor), or put an empty at the world origin in the Mirror Object field of the Mirror panel.

5. Changing one number: the power of an unapplied stack

Starting point: continue from the Fence_Picket with Count 6 from section 3.

Do this:

  1. Double-click the Count field in the Array panel of the Modifiers tab.
  2. Type 4 and press Enter.

Count in the Array panel changed to 4, and the pickets in the viewport reduced to eight, four on each side. Box 1 on the Count field Figure 5. The result of Count 4. The pickets drop to eight right away (Blender screenshot)

The pickets change to four on each side, eight in all, right away. After checking, set Count back to 6.

Do this: look at the bevel values.

  1. Click the small arrows at the left of the Array and Mirror panel headers to fold them.
  2. In the Bevel panel below, read Amount, Segments, and Limit Method.

The Modifiers tab with the Array and Mirror panels folded and only the Bevel panel open. Box 1 on Width Type Offset, Amount 0.008 m, Segments 3, Limit Method Angle, and Angle 30° Figure 6. Folded Array and Mirror, open Bevel: 1) Amount 0.008 m, Segments 3, Limit Method Angle (Blender screenshot)

Amount is the width of the corner cut, and Segments is how many steps the rounded face has. With 1 segment you get one slanted face; with 3 it looks rounded. With Limit Method set to Angle, only edges bent more than 30° are cut. The picket is 0.1 m wide, so the AI kept Amount small at 0.008 m. From a distance the rounding is hard to see; if you want it rounder, try raising Amount to 0.02.

6. Writing a better question

The count difference in this run and the common mistakes in section 4 can be avoided from the start by adding one line to the question.

ProblemWhat to add to the question (example)
The count is read differently"6 to the right including the original, 12 in all"
The bevel is uneven"Make the size by moving points, not with Scale, or apply the scale after changing it"
Mirror overlaps the middle"Make the whole fence symmetric around the world origin"
You end up trusting words only"When done, tell me the picket count and spacing as numbers"

When menu names differ: the Add Modifier > Generate menu in Blender 5.2 has two Arrays. Array at the top is the new node-based array, and Array (Legacy) right below it is the older array the AI added in this guide. The two panels have different fields. When you add one by hand to follow along, choose Array (Legacy) to get the same panel as in Figure 1.

A Modifiers tab with two arrays added in turn to one object. Box 1 is the new Array with Shape, Count, Offset Method, and Offset, Rotation, Scale fields. Box 2 is Array (Legacy) with Fit Type, Count, Relative Offset, and Factor X fields Figure 7. Two arrays added from the menu: 1) the new Array (nodes), 2) Array (Legacy). The image shows the second array named Array.001 because the names clashed (Blender screenshot)

When you ask AI, naming which one also avoids confusion. Example: "with the Array (Legacy) modifier".

Practice

  1. Change Count to 3 and count how many pickets you get.
  2. Open Constant Offset and change Distance X from 0.25 to 0.4. Write down how wide the gap between pickets becomes (a picket is 0.1 m wide).
  3. Change Segments in Bevel to 1 and look closely at how the corners change.
  4. Press Tab to enter Edit Mode, press A to select all points, and move the picket toward the origin with G → X → -0.125 → Enter. Check whether the middle pickets overlap as in mistake 2 of section 4, then undo with Ctrl+Z.

Common mistakes and fixes

SymptomCauseFix
There are two more pickets than expected"6 to the right" was read as the original plus 6Set Array Count to 6. Write the total, such as "12 in all," in the question
One picket seems to be missingThe Mirror origin is in the middle of a picket, so two overlapMove the origin to the world origin or use Mirror Object
The bevel is round in one direction onlyScale is not appliedCtrl+A > Scale
Changing Count does not change the shapeThe AI applied the modifiersAdd "do not apply" to the question and ask again. Saving before makes it easy to go back
An Array panel added by hand looks different from the imageThe new Array (nodes) was chosenAdd Modifier > Generate > Array (Legacy)
The rounded corners are hard to seeBevel Amount is small for the picket sizeRaise Amount a little at a time

Summary

Ask "do not apply, just stack" → read the answer to the last line → check count, scale, and origin on screen and in numbers → refine with panel numbers. In the next guide, Letting AI set up lights and materials, you hand three-point lighting and PBR materials to AI the same way.

Further reading

References

KEEP EXPLORING

Read next

EEVEE render of a grey fighter-jet model sitting at an angle on a beige floor, with a shadow to the lower rightBlender
Blender and AI11 min read

Cleaning up an AI-generated 3D model in Blender: merge split points, set the size, reduce faces

Import a fighter-jet model GLB made with Meshy into Blender 5.2, read its state from the statistics, merge split points with Merge by Distance, tidy normals and smooth shading, set the size, and cut the face count to one tenth with Decimate.

A PowerShell window. Running blender-mcp through uv with --transport http --port 9191 prints Started server process, StreamableHTTP session manager started, and Uvicorn running on http://127.0.0.1:9191 (terminal screenshot)Blender
Blender and AI11 min read

Using Blender MCP with ChatGPT: custom MCP servers in ChatGPT on the web, secure tunnels, and how Codex differs

What to know when you want to use Blender MCP with GPT: why the official Blender MCP cannot plug straight into ChatGPT on the web, the two routes OpenAI's docs describe (a public HTTPS address or Secure MCP Tunnel), and how they compare with Codex, the easiest way to drive Blender with a GPT model.

A 960x540 EEVEE render. A red cube on a gray floor with a smooth gray monkey head resting on it by the tip of its chin, and a shadow falling back to the left (Blender render)Blender
Blender and AI15 min read

Ten first prompts for Blender MCP: from reading the scene to rendering, seen through the real answers

After connecting the official Blender MCP to an AI app, we sent ten first prompts in order: read, create, edit, render, and review. This guide shows what the AI did for each one and what a person still needs to check, with screenshots.

EEVEE render of a bright gold metal monkey head on a grey floor, with sharp reflections on the brows, eyeballs, and ears and an interior photo background with beige wallpaper and a dark window (Blender render)Blender
Blender and AI14 min read

Letting AI set up lights and materials: three-point lighting and PBR over MCP, compared in renders

Through the official Blender MCP, hand the three-point lighting and the gold metal and matte floor materials of a monkey head scene to AI, read the real result in the panels, and compare three EEVEE renders (default light, the AI's result, and a hand-tuned version) to learn why metal looks dull and which values are worth changing by hand.

EEVEE render of a low-poly room with two walls and a wooden floor. A bookshelf with books on its shelves at front left, a bed with a blue blanket, a pinkish rug, a desk, chair, and desk lamp under the window, and a potted plant in the right corner (Blender render)Blender
Blender and AI13 min read

Building a low-poly room with Blender MCP: the order of questions and what to fix by hand

Send five questions in one conversation through the official Blender MCP to build a low-poly room with walls, a window, furniture, props, and lights, then use the AI's notes, a check that does not change the scene, and the first render to fix the bookshelf direction, the default light left on, and a dim desk lamp by hand.

Two bar charts. Left: tokens for nine runs (A1 to C3), mostly input read from cache. Right: estimated cost, highest for the first run of each group and lowest for C2 and C3 at 0.065 dollars (chart)Blender
Blender and AI12 min read

Measuring Blender MCP token use: the same job three times each, and how to save

With Claude Code and the official Blender MCP, we ran 'place five cylinders' in three question styles, three times each, nine runs in all. We measured input and output tokens, time, and estimated cost, and compared with real numbers how much it helps to skip screenshot checks and to give exact values.