Blender MCP
Provides tools for interacting with Blender scenes and workflows, including scene inspection, object and mesh modeling, materials, modifiers, cameras, lights, Geometry Nodes, animation keyframes, rendering, viewport screenshots, import/export, and project save/load.
Supports Sketchfab as an optional asset provider, allowing access to 3D assets within Blender workflows.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Blender MCP add a red cube, a camera, and a light, then render a preview"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Blender MCP
Production-oriented Model Context Protocol (MCP) integration for Blender. AI clients (Claude Desktop, Cursor, VS Code, Codex, Windsurf, and others) can inspect scenes, model, shade, light, render, and visually verify work through a local bridge.
This project is not affiliated with the Blender Foundation.
Features
MCP stdio server + Blender addon TCP bridge (localhost by default)
Scene inspection, object/mesh modeling, materials, modifiers, cameras, lights
Geometry Nodes helpers with socket inspection
Viewport screenshots and render previews for visual verification
Animation keyframes, import/export, project save/load
Optional asset providers (Poly Haven, Sketchfab, Poly Pizza, AI 3D stubs)
Script execution with explicit permission modes (restricted / confirmation / trusted)
Auth token, path allow-lists, render limits, tool-group toggles
Related MCP server: SandBase MCP
Prerequisites
uv (required)
Blender 3.6+ (4.x recommended)
Quick start
Install site (Astro / GitHub Pages): axcel-blade.github.io/Blender-MCP
Windows (PowerShell) — one liner:
irm https://axcel-blade.github.io/Blender-MCP/install.ps1 | iexmacOS / Linux:
curl -LsSf https://axcel-blade.github.io/Blender-MCP/install.sh | shOr from a local clone with uv:
uv sync
uv run blender-mcp install-addonBlender → Edit → Preferences → Add-ons → enable Blender MCP
3D Viewport → press N → AI / MCP → Start
Point your MCP client at:
{
"mcpServers": {
"blender": {
"command": "uv",
"args": ["run", "--directory", "C:/path/to/Blender-MCP", "blender-mcp"]
}
}
}Documentation
Document | Purpose |
System design | |
Install & MCP client setup | |
Config & environment | |
MCP tools overview | |
Technical security controls | |
Common failures | |
Version history | |
Git Flow & contribution guide | |
How to get help | |
Vulnerability reporting | |
Community standards | |
MIT license |
Architecture
AI Client --stdio MCP--> blender-mcp server --TCP NDJSON--> Blender addon --bpy--> BlenderSee docs/ARCHITECTURE.md.
Git Flow
Branch | Purpose |
| Production releases |
| Integration |
| Features → |
| Release prep → |
| Urgent fixes from |
Details: CONTRIBUTING.md.
Security
Bridge binds to
127.0.0.1unless remote mode is explicitly enabledOptional shared auth token
execute_blender_codeis not a sandbox — keepscript_permission=restrictedunless you trust the clientPath allow-lists for file operations when configured
Report vulnerabilities privately via SECURITY.md.
Development
This project is managed exclusively with uv (lockfile + virtualenv).
uv sync
uv run pytest
uv run ruff check src tests
uv run pyrightDiagnose a live Blender connection:
uv run blender-mcp diagnoseLicense
MIT © AXCEL BLADE — see LICENSE.
External asset providers retain their own licenses and attribution requirements.
Available Tools
156 toolsadd_geometry_nodeC
Add a node to a geometry node group.
Inputs: node_type (required), group_name (required), node_name. Returns: object
add_modifierC
Add a modifier to an object.
Inputs: name (required), modifier_type (required), settings, modifier_name. Returns: object
animate_locationC
Keyframe object location.
Inputs: name (required), frame (required), location (required). Returns: object
animate_rotationC
Keyframe object rotation.
Inputs: name (required), frame (required), rotation (required). Returns: object
animate_scaleD
Keyframe object scale.
Inputs: name (required), frame (required), scale (required). Returns: object
apply_modifierD
Apply a modifier.
Inputs: name (required), modifier_name (required). Returns: object
apply_transformD
Apply object transform.
Inputs: name (required), scale, location, rotation. Returns: object
assign_materialC
Assign a material to an object.
Inputs: object_name (required), material_name (required), slot. Returns: object
assign_textureC
Assign an image texture to a material input.
Inputs: image_path (required), material_name (required), input_name. Returns: object
bevel_meshC
Bevel mesh edges.
Inputs: name (required), amount, segments. Returns: object
blender_mcp_versionA
Return Blender MCP server version and bridge configuration (no secrets).
Inputs: no named parameters.
camera_viewC
Return camera framing metadata for visual verification.
Inputs: name. Returns: object
capture_camera_viewC
Capture from the active or named camera.
Inputs: camera, max_size. Returns: object
capture_material_previewC
Capture a material preview viewport image.
Inputs: max_size. Returns: object
capture_multiple_anglesC
Capture multiple orthographic/perspective angles.
Inputs: angles, max_size. Returns: object
capture_renderC
Capture or return the latest render preview.
Inputs: max_size. Returns: object
capture_scene_previewC
Capture a scene preview image.
Inputs: max_size. Returns: object
capture_viewportC
Capture the viewport with optional view/shading overrides.
Inputs: view, shading, max_size. Returns: object
capture_wireframeC
Capture a wireframe viewport image.
Inputs: max_size. Returns: object
configure_camera_depth_of_fieldD
Configure camera depth of field.
Inputs: name (required), fstop, enabled, focus_object, +1 more. Returns: object
configure_camera_lensC
Configure camera lens settings.
Inputs: name (required), lens, type, ortho_scale. Returns: object
connect_geometry_nodesC
Connect two geometry nodes.
Inputs: to_node (required), from_node (required), to_socket (required), group_name (required), +1 more. Returns: object
create_animation_previewC
Capture frames for an animation preview strip.
Inputs: step, frame_end, frame_start. Returns: object
create_array_of_objectsC
Duplicate an object into an array along an offset.
Inputs: name (required), count, offset. Returns: object
create_basic_shaderC
Create a basic Principled material.
Inputs: name, color. Returns: object
create_cameraD
Create a camera.
Inputs: lens, name, location, rotation, +1 more. Returns: object
create_circleD
Create a circle.
Inputs: name, radius, location, vertices, +1 more. Returns: object
create_collectionD
Create a collection.
Inputs: name (required), parent. Returns: object
create_coneD
Create a cone.
Inputs: name, depth, radius1, radius2, +2 more. Returns: object
create_cubeC
Create a cube.
Inputs: name, size, location, collection. Returns: object
create_curveC
Create a Bezier curve.
Inputs: name, location, collection. Returns: object
create_cylinderD
Create a cylinder.
Inputs: name, depth, radius, location, +2 more. Returns: object
create_emission_materialC
Create an emission material.
Inputs: name, color, strength. Returns: object
create_emptyD
Create an empty.
Inputs: name, location, collection, empty_display_type. Returns: object
create_geometry_nodes_modifierC
Add a Geometry Nodes modifier.
Inputs: name (required), group_name. Returns: object
create_glass_materialD
Create a glass material.
Inputs: ior, name, roughness. Returns: object
create_hierarchyC
Create nested collections by path names.
Inputs: names (required). Returns: object
create_ico_sphereD
Create an icosphere.
Inputs: name, radius, location, collection, +1 more. Returns: object
create_keyframeC
Insert a keyframe on an object.
Inputs: name (required), data_path (required), frame, index. Returns: object
create_lightD
Create a light.
Inputs: name, color, energy, location, +1 more. Returns: object
create_lighting_setupC
Create a lighting preset (three_point, studio, product, outdoor, interior, cinematic).
Inputs: preset, target. Returns: object
create_materialC
Create a Principled BSDF material.
Inputs: name, metallic, roughness, base_color. Returns: object
create_meshC
Create an empty mesh object.
Inputs: name, location. Returns: object
create_metallic_materialD
Create a metallic material.
Inputs: name, color, roughness. Returns: object
create_node_groupC
Create a geometry node group.
Inputs: name. Returns: object
create_objectD
Create a generic object by type.
Inputs: object_type (required), name, scale, location, +2 more. Returns: object
create_planeD
Create a plane.
Inputs: name, size, location, collection. Returns: object
create_primitiveC
Create a mesh primitive (cube, uv_sphere, cylinder, ...).
Inputs: primitive (required), name, size, scale, +4 more. Returns: object
create_sphereD
Create a UV sphere.
Inputs: name, radius, location, collection. Returns: object
create_textC
Create a text object.
Inputs: name, text, location, collection. Returns: object
create_torusD
Create a torus.
Inputs: name, location, collection, major_radius, +1 more. Returns: object
create_transparent_materialC
Create a transparent material.
Inputs: name, alpha, color. Returns: object
create_uv_sphereC
Create a UV sphere with segment/ring control.
Inputs: name, rings, radius, location, +2 more. Returns: object
decimate_meshD
Decimate a mesh.
Inputs: name (required), ratio. Returns: object
delete_collectionD
Delete a collection.
Inputs: name (required), confirmed. Returns: object
delete_geometryC
Delete mesh geometry components.
Inputs: name (required), mode. Returns: object
delete_lightC
Delete a light object.
Inputs: name (required), confirmed. Returns: object
delete_objectB
Delete a single object (requires confirmed=true when confirm_destructive is on).
Inputs: name (required), confirmed. Returns: object
delete_objectsC
Delete multiple objects by explicit names.
Inputs: names (required), confirmed. Returns: object
diagnose_blender_connectionB
Diagnose MCP-to-Blender connectivity and capabilities.
Inputs: no named parameters. Returns: object
duplicate_objectD
Duplicate an object.
Inputs: name (required), linked, new_name. Returns: object
edit_meshC
Run a structured mesh edit operation on an object.
Inputs: name (required), operation (required), params. Returns: object
execute_blender_codeC
Execute Python code inside Blender when script permissions allow it.
Warning: this is not a secure sandbox. It can access files, network, and
system capabilities available to the Blender process.
Inputs: code (required), timeout, confirmed, preview_only. Returns: object
export_assetC
Export selected or named objects.
Inputs: path (required), objects, overwrite. Returns: object
export_sceneC
Export the full scene.
Inputs: path (required), overwrite. Returns: object
export_selected_objectsC
Export currently selected objects.
Inputs: path (required), overwrite. Returns: object
extrude_meshC
Extrude selected mesh geometry (object-mode helper).
Inputs: name (required), offset. Returns: object
fill_facesC
Fill selected boundary with faces where possible.
Inputs: name (required). Returns: object
focus_camera_on_objectC
Point/focus a camera at an object.
Inputs: camera (required), object_name (required). Returns: object
frame_objectsC
Frame objects in a camera view.
Inputs: camera, object_names. Returns: object
get_active_objectA
Return the active object, if any.
Inputs: no named parameters. Returns: object
get_animation_infoC
Summarize animation data and timeline.
Inputs: no named parameters. Returns: object
get_camera_infoC
Get camera information.
Inputs: name. Returns: object
get_collection_infoC
Get collection details.
Inputs: name (required). Returns: object
get_collectionsB
List collections in the scene.
Inputs: no named parameters. Returns: object
get_current_frameA
Get the current frame number.
Inputs: no named parameters. Returns: object
get_light_infoD
Get light information.
Inputs: name. Returns: object
get_lightsA
List lights in the scene.
Inputs: no named parameters. Returns: object
get_material_infoC
Inspect a material and its nodes.
Inputs: name (required). Returns: object
get_material_nodesC
List nodes in a material node tree.
Inputs: name (required). Returns: object
get_materialsB
List materials in the blend file.
Inputs: no named parameters. Returns: object
get_mesh_statisticsB
Get mesh statistics for an object or the whole scene.
Inputs: name. Returns: object
get_modifiersC
List modifiers on an object.
Inputs: name (required). Returns: object
get_node_socket_infoC
Get socket schema for a node.
Inputs: node_name (required), group_name (required). Returns: object
get_object_infoB
Get detailed info for a single object by name.
Inputs: name (required). Returns: object
get_objectsC
List objects with optional filtering and pagination.
Inputs: limit, offset, name_filter, object_type. Returns: object
get_object_transformC
Get location/rotation/scale for an object.
Inputs: name (required), space. Returns: object
get_project_pathB
Get the current blend file path.
Inputs: no named parameters. Returns: object
get_render_resultA
Return the last render result path/metadata.
Inputs: no named parameters. Returns: object
get_render_settingsC
Get render settings.
Inputs: no named parameters. Returns: object
get_scene_hierarchyB
Return the scene collection hierarchy.
Inputs: no named parameters. Returns: object
get_scene_infoC
Return detailed information about the current Blender scene.
Inputs: no named parameters. Returns: object
get_scene_settingsB
Get unit/frame/gravity and related scene settings.
Inputs: no named parameters. Returns: object
get_scene_summaryC
Return a compact summary of the current scene.
Inputs: no named parameters. Returns: object
get_selected_objectsB
List currently selected objects.
Inputs: no named parameters. Returns: object
get_viewport_screenshotC
Capture the 3D viewport as an image for visual verification.
Inputs: shading, max_size. Returns: object
import_assetC
Import a 3D asset (FBX/OBJ/STL/GLTF/GLB/USD/BLEND).
Inputs: path (required), collection. Returns: object
import_external_assetC
Download metadata/URLs then ask Blender to import when a local path is available.
Inputs: source (required), asset_id (required), asset_type, resolution. Returns: object
import_polyhaven_via_blenderB
Ask Blender to import a Poly Haven asset by id (addon-side download).
Inputs: asset_id (required), asset_type, resolution. Returns: object
inset_facesC
Inset faces on a mesh object.
Inputs: name (required), thickness. Returns: object
inspect_animationD
Inspect animation data.
Inputs: name. Returns: object
inspect_node_groupC
Inspect a node group structure.
Inputs: group_name (required). Returns: object
inspect_renderA
Return metadata about the latest captured/rendered image.
Inputs: no named parameters. Returns: object
join_objectsC
Join multiple mesh objects.
Inputs: names (required), target. Returns: object
list_asset_providersA
List configured asset providers and credential status (no secrets).
Inputs: no named parameters. Returns: object
list_geometry_node_typesA
List available geometry node type identifiers.
Inputs: no named parameters. Returns: object
list_materialsC
List all materials.
Inputs: no named parameters. Returns: object
list_modifiersC
List modifiers on an object with settings.
Inputs: name (required). Returns: object
merge_verticesC
Merge vertices by distance.
Inputs: name (required), distance. Returns: object
mesh_cleanupD
Run common mesh cleanup operations.
Inputs: name (required). Returns: object
modify_cameraD
Modify camera settings.
Inputs: name (required), settings (required). Returns: object
modify_lightD
Modify light settings.
Inputs: name (required), settings (required). Returns: object
modify_materialC
Modify Principled BSDF inputs on a material.
Inputs: name (required), settings (required). Returns: object
modify_modifierD
Update modifier settings.
Inputs: name (required), settings (required), modifier_name (required). Returns: object
modify_objectC
Modify common object properties.
Inputs: name (required), scale, location, rotation, +1 more. Returns: object
move_objectD
Move an object.
Inputs: name (required), location (required), space. Returns: object
move_object_to_collectionC
Move an object into a collection.
Inputs: collection (required), object_name (required), exclusive. Returns: object
open_blend_fileD
Open a .blend file.
Inputs: path (required). Returns: object
organize_sceneC
Organize objects into logical collections.
Inputs: structure. Returns: object
recalculate_normalsC
Recalculate mesh normals.
Inputs: name (required), inside. Returns: object
remesh_meshC
Apply remesh via modifier and apply it.
Inputs: name (required), mode, octree_depth. Returns: object
remove_keyframeC
Remove a keyframe.
Inputs: name (required), data_path (required), frame. Returns: object
remove_materialC
Remove a material slot from an object.
Inputs: object_name (required), slot. Returns: object
remove_modifierC
Remove a modifier from an object.
Inputs: name (required), modifier_name (required). Returns: object
rename_collectionD
Rename a collection.
Inputs: name (required), new_name (required). Returns: object
rename_objectC
Rename an object.
Inputs: name (required), new_name (required). Returns: object
render_animationC
Render an animation range.
Inputs: frame_end, frame_start, output_path. Returns: object
render_finalC
Render a higher-quality still.
Inputs: samples, output_path. Returns: object
render_previewC
Render a fast preview image.
Inputs: samples, resolution_percentage. Returns: object
render_sceneD
Render the current frame.
Inputs: engine, samples, output_path, write_still, +1 more. Returns: object
reset_transformC
Reset object transform channels.
Inputs: name (required), scale, location, rotation. Returns: object
rotate_objectC
Rotate an object (Euler radians).
Inputs: name (required), rotation (required), space. Returns: object
save_asC
Save the blend file to a new path.
Inputs: path (required), overwrite. Returns: object
save_blend_fileC
Save the current .blend file.
Inputs: path, overwrite. Returns: object
scale_objectD
Scale an object.
Inputs: name (required), scale (required). Returns: object
search_assetsC
Search external asset providers (polyhaven, sketchfab, polypizza).
Inputs: query (required), source (required), limit, asset_type. Returns: object
separate_objectsC
Separate a mesh object.
Inputs: name (required), mode. Returns: object
set_active_cameraC
Set the active scene camera.
Inputs: name (required). Returns: object
set_animation_rangeC
Set scene animation range.
Inputs: frame_end (required), frame_start (required). Returns: object
set_current_frameC
Set the current frame.
Inputs: frame (required). Returns: object
set_geometry_node_inputC
Set a geometry node input value.
Inputs: value (required), node_name (required), group_name (required), input_name (required). Returns: object
set_keyframeC
Set a property value and keyframe it.
Inputs: name (required), value (required), data_path (required), frame. Returns: object
set_material_colorC
Set material base color.
Inputs: name (required), color (required). Returns: object
set_material_nodesC
Apply structured node operations to a material.
Inputs: name (required), operations (required). Returns: object
set_material_propertiesC
Set common Principled BSDF properties.
Inputs: name (required), properties (required). Returns: object
set_object_originC
Set object origin.
Inputs: name (required), origin_type. Returns: object
set_object_parentD
Parent an object.
Inputs: name (required), parent (required), keep_transform. Returns: object
set_object_shadingC
Set smooth or flat shading.
Inputs: name (required), smooth. Returns: object
set_object_transformC
Set full object transform.
Inputs: name (required), scale, location, rotation. Returns: object
set_object_visibilityC
Set object visibility flags.
Inputs: name (required), hide_render, hide_select, hide_viewport. Returns: object
set_output_pathC
Set render output path.
Inputs: path (required). Returns: object
set_render_settingsC
Update render settings with configured limits enforced.
Inputs: settings (required). Returns: object
smooth_meshC
Smooth mesh vertices.
Inputs: name (required), factor, repeat. Returns: object
subdivide_meshC
Subdivide mesh faces.
Inputs: name (required), cuts. Returns: object
triangulate_meshD
Triangulate a mesh.
Inputs: name (required). Returns: object
unparent_objectC
Unparent an object.
Inputs: name (required), keep_transform. Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 156 tools
Many tools overlap heavily: get_materials vs list_materials, get_modifiers vs list_modifiers, get_scene_info vs get_scene_summary, and create_primitive vs dozens of primitive-specific creators. Render/capture tools also blur together (render_scene, render_preview, render_final; capture_viewport, get_viewport_screenshot, capture_scene_preview). An agent can still pick a reasonable tool in many cases, but several boundaries are genuinely unclear.
The surface is overwhelmingly snake_case, but verb conventions are inconsistent: get_ and list_ are used interchangeably for the same resource (materials/modifiers), some tools are noun-only (camera_view, mesh_cleanup, organize_scene), and capture_ vs render_ vs get_ overlap conceptually. It remains readable, but the naming does not reliably encode resource+action.
156 tools is far beyond a well-scoped surface, even for a broad Blender domain. The set contains many convenience wrappers and near-duplicates (create_sphere, create_uv_sphere, create_primitive, create_cube, etc.) that inflate the count without adding distinct capability.
Coverage is extremely broad: scene/object inspection, mesh editing, materials and nodes, cameras, lights, rendering, animation, assets, file operations, collections, viewport capture, and Python execution. Almost every core Blender workflow appears represented, with no obvious lifecycle dead ends.
Maintenance
Related MCP Connectors
Connect your AI assistant to Blender for 3D modeling, Geometry Nodes, materials, lighting, animation, rendering and export. Inspect real scenes and viewport images, refine the result and restore checkpoints. OAuth connects your chat to Blender Bridge Desktop on an online Windows 10/11 x64 computer running Blender 4.2+. The chat can run on another device. Requires an active Blender Bridge account; 7-day trial available. AI model access is separate.
Secure AI access to Windows, macOS, Linux computers and a hosted Cloud Browser.
Connect AI assistants to your approved computer for files, commands, browser and desktop tasks.
On-brand creative studio for AI agents: images, video, audio, and 3D.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables MCP-capable AI clients to control a local Blender workstation through a secure HTTP gateway for 3D scene inspection, modeling, rigging, animation, materials, rendering, import/export, and controlled Python automation. Supports permission profiles and authentication modes to keep Blender execution local while allowing AI-assisted production workflows.1MIT
- AlicenseNot gradedqualityAmaintenanceEnables AI clients to discover, inspect, and run 2,000+ AI models and data APIs—including LLMs, image, video, and audio generation—through a single authenticated local bridge with per-call pricing shown before execution. It also supports run status tracking, account balance checks, and rollback from one browser sign-in.556,724 npmMIT
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to build models, import assets, animate scenes, render images, and save versioned projects through validated MCP tools backed by headless Blender, with a live browser viewer for inspection.GPL 3.0
- AlicenseCqualityBmaintenanceEnables AI clients to inspect, visualize, modify, launch, and manage Autodesk BIM models, views, families, and documents through a local authenticated bridge, including direct code execution.33MIT