blender-mcp
Create, iterate, and export 3D scenes in Blender via MCP with visual feedback loops.
Skill by uset82
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/uset82-blender-mcp
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: design, data, agent-building
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/blender-mcp
What's inside
- Skill name
blender-mcp- Description
- Create, iterate, and export 3D scenes in Blender via MCP with visual feedback loops. Use for product visualization, interior concepts, motion design, hero renders, or when StackSelection includes the blender-mcp integration. Requires a running Blender instance with the MCP addon enabled.
- License
- MIT
- Instructions
- 134 lines in SKILL.md · view on GitHub
Files (5)
- SKILL.md
- claude.yaml · agents/
- openai.yaml · agents/
- codex-gpt-blender-mcp.md · docs/
- integration-img2threejs.md · docs/
SKILL.md (first lines)
# blender-mcp — Blender Model Context Protocol Integration Build, render, inspect, and fix 3D scenes in Blender through MCP tool calls with a visual self-correction loop. The agent describes the scene, Blender creates it, the agent sees the render, and corrections continue until the result matches the original intent. **Primary path in WebDesigner: Codex / GPT-5.4** (vision + coding + tool use), routed via `skillOverrides.blender-mcp` in `.antigravity/runtime/routing-policy.json`. Still host-portable: Claude Code, Codex IDE, OpenCode, or WebDesigner can run the same MCP loop. Wherever this doc says "viewport screenshot" or "agent vision", use whatever visual feedback the MCP server provides. **Narrative / product essay:** [docs/codex-gpt-blender-mcp.md](./docs/codex-gpt-blender-mcp.md) ## WebDesigner Contract - **Stage**: `design` (3D concept development), `build` (scene creation and iteration) - **Reads**: `TaskIntent`, `StackSelection`, design brief, reference images, generated workspace - **Emits artifacts**: `blender-scene-spec`, `blender-render`, `blender-review-log`, `exported-asset` - **MCP dependency**: `blender` MCP server (via `uvx blender-mcp` or local addon) - **Activation**: `StackSelection.integrations` contains `blender-mcp`, or `TaskIntent.constraints.requiresBlenderMCP` is true, or the user explicitly asks for Blender scene creation ### WebDesigner Rules - Require a running Blender instance with the MCP addon connected on `localhost:9876`. - Use only project-scoped MCP tools: `get_scene_info`, `get_object_info`, `get_screenshot`, `execute_blender_code`. - Treat `execute_blender_code` as arbitrary local code execution; respect approval prompts when configured. - Keep PolyHaven, Sketchfab, Hyper3D, Hunyuan, and other network-backed integrations disabled per project policy. - Export GLB/PNG assets to `.codex-avatar/exports/blender/` or the generated workspace's asset directory. - Work on explicit copies; never overwrite user-selected source scenes. - Record render screenshots, comparison sheets, and export paths in the `ArtifactManifest`. ### WebDesigner Process 1. Confirm `integrations` contains `blender-mcp` or user explicitly requested Blender scene work. 2. Verify Blender MCP connection via `get_scene_info`. 3. Parse the user description into a scene specification (objects, materials, lighting, camera). 4. Follow the iterative build loop: create → render → inspect → fix. 5. Export final assets (GLB, PNG, optional blend) to the generated workspace. 6. Emit `blender-scene-spec`, `blender-render`, and `blender-review-log` artifacts. ## When To Use - Product visualization: bottles, electronics, furniture, fictional devices - Interior concepts: room layouts, furniture arrangements, lighting studies - Motion design: procedural forms, extruded typography, geometric loops - Hero renders: landing page backgrounds, marketing materials - Asset creation: 3D props for web/game projects, WebGL/Three.js source assets - Prototyping: quick 3D mockups before committing to detailed procedural code ## Core Promise Build from description in stages — never attempt a complete scene in one prompt: 1. **Composition first**: Block out main objects, camera angle, basic layout. 2. **Materials and lighting**: Apply initial materials, establish lighting rig. 3. **Render and inspect**: Capture viewport/render, identify visible mismatches. 4. **Targeted corrections**: Adjust specific settings (camera, lighting, materials) based on inspection. 5. **Iterate**: Repeat render-inspect-fix until the result matches intent. State explicitly when output is approximate or stylized. A single description cannot specify every detail — ask for clarification instead of making arbitrary choices. ## Required Inputs - Scene description (objects, mood, purpose, camera angle preferences) - Output format: product shot, interior render, animated loop, exportable asset - Quality level: quick prototype, polished render, production asset ## The Loop (MCP tools do execution; agent vision does judgment) ### Available MCP Tools | Tool | Purpose | |------|---------| | `mcp__blender__get_scene_info` | Inspect current scene state | | `mcp__blender__get_object_info` | Get details about specific objects | | `mcp__blender__get_screenshot` | Capture viewport for visual feedback | | `mcp__blender__execute_blender_code` | Run Python code against Blender's bpy API | ### Iteration Process 1. **Scene setup**: Create base geometry with `execute_blender_code`: ```python import bpy bpy.ops.mesh.primitive_cylinder_add(radius=0.7, depth=1.6) bpy.ops.object.light_add(type="AREA", location=(4, -4, 6)) bpy.context.object.data.energy = 800 bpy.ops.object.camera_add(location=(5, -6, 3)) bpy.context.scene.camera = bpy.context.object ``` 2. **Render capture**: Use `get_screenshot` to see current state. 3. **Visual inspection**: Analyze the screenshot with agent vision: - Is the subject framed correctly? - Are materials reading as intended? - Is lighting flattering or problematic? - Does composition match the description? 4. **Correction**: Issue targeted fixes based on inspection: - Camera too close? Increase distance, decrease focal length. - Harsh shadows? Soften light, adjust angle. - Material too shiny? Reduce roughness, check reflection. - Object hidden? Adjust position, check visibility. 5. **Record review**: Log the iteration with scores and decision: - `continue`: Result matches intent, proceed to next stage. - `refine-scene`: Adjustments needed, continue iteration. - `request-input`: Need clarification from user. - `stop`: Cannot achieve requested result, explain limitation. ## Gates (do not skip) - **Connection check**: Verify Blender MCP is reachable before scene work. - **Screenshot feedback**: Corrections must be based on actual render output, not assumptions. - **User approval**: Major direction changes require confirmation. - **Export validation**: Verify exported assets are valid (GLB header, PNG signature, file size). ## Limitations
From uset82/webdesigner, MIT.
More from uset82/webdesigner
- 0.524score
friction-vector
uset82
Design and build high-fidelity 2D vector animations, morphing paths, animated SVGs (SMIL/CSS), and interactive Skia-inspired Canvas components using the…
SkillBFreshness: BClaude Code +9 - 0.491score
frontend-design
uset82
Guidance for distinctive, intentional visual design when building new UI or reshaping an existing one.
SkillBFreshness: BClaude Code +9 - 0.474score
img2threejs
uset82
Turn an object or character reference image into a quality-gated, animation-ready procedural Three.js model built in code.
SkillCFreshness: CClaude Code +9 - 0.471score
framework-selector
uset82
Produces a layered stack selection for the curated WebDesigner v1 support matrix.
SkillCFreshness: CClaude Code +9 - 0.471score
security-audit
uset82
Runs the dedicated security stage for generated workspaces.
SkillCFreshness: CClaude Code +9 - 0.469score
animate-ui
uset82
Add and adapt imskyleen/animate-ui registry components for intentional animated React interfaces.
SkillCFreshness: CClaude Code +9 - 0.469score
gsap-animation
uset82
Add GSAP (GreenSock Animation Platform) timelines, scroll-driven effects, SVG morphing, text split animations, smooth scrolling, draggable interactions, and…
SkillCFreshness: CClaude Code +9 - 0.469score
code-generator
uset82
Converts approved design artifacts into idiomatic code inside the generated workspace selected by the plan stage.
SkillCFreshness: CClaude Code +9 - 0.469score
project-scaffolder
uset82
Creates the generated workspace for a selected stack.
SkillCFreshness: CClaude Code +9 - 0.467score
frontend-skill
uset82
Steers frontend design and code implementation toward visually distinctive, premium interfaces.
SkillCFreshness: CClaude Code +9 - 0.467score
deploy-advisor
uset82
Selects a deployment target for the chosen stack and emits deployable configuration artifacts for the generated workspace.
SkillCFreshness: CClaude Code +9 - 0.467score
remotion-video-generator
uset82
Programmatically creates React-based animations and video compositions, rendering them into MP4, WebP, or GIF formats using Remotion.
SkillCFreshness: CClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 1
- Last commit (this folder)
- Jul 22, 2026 (3 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/uset82/webdesigner
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from uset82
- 0.524score
friction-vector
uset82
Design and build high-fidelity 2D vector animations, morphing paths, animated SVGs (SMIL/CSS), and interactive Skia-inspired Canvas components using the…
SkillBFreshness: BClaude Code +9 - 0.491score
frontend-design
uset82
Guidance for distinctive, intentional visual design when building new UI or reshaping an existing one.
SkillBFreshness: BClaude Code +9 - 0.474score
img2threejs
uset82
Turn an object or character reference image into a quality-gated, animation-ready procedural Three.js model built in code.
SkillCFreshness: CClaude Code +9 - 0.471score
security-audit
uset82
Runs the dedicated security stage for generated workspaces.
SkillCFreshness: CClaude Code +9 - 0.471score
framework-selector
uset82
Produces a layered stack selection for the curated WebDesigner v1 support matrix.
SkillCFreshness: CClaude Code +9 - 0.469score
animate-ui
uset82
Add and adapt imskyleen/animate-ui registry components for intentional animated React interfaces.
SkillCFreshness: CClaude Code +9
Cite this page
blender-mcp skill. Agent Mart. https://agentmart.dev/skills/uset82-blender-mcp (last updated 2026-10-24)