- Home
- Skills
- Game Development
- BlendPro: Blender for AI Agents
More screenshots
Works with the AI Agents you already use
BlendPro: Blender for AI Agents
Most agent-driven Blender work runs code, takes a screenshot to see what happened, and repeats.
Free
BlendPro: Blender for AI Agents
Example session with this skill installed
Using BlendPro, build a stylised wooden treasure chest with iron straps, rivets and a pile of gold coins. Texture it, check it against the Quest budget and export a GLB for WebXR. Save a render of each stage.
- Read your context and instructions
- Compiled the blendpro
Summary
The chest is built, textured, checked and exported. Everything ran in a headless Blender 5.2, and the only images the agent opened were final renders.
Stages
| Stage | How it was done |
|---|---|
| Blockout | One run_python call: hollow body, hinged lid, straps, rim band, lock, handles, 16 rivets (linked duplicates), gold pile, 70 coins, pedestal, camera rig |
| Cleanup | mesh_analysis reported 196 duplicate vertices, degenerate faces, n-gons and an unapplied scale. The mesh_cleanup, fix_ngons and apply_transforms recipes fixed them, and a second report came back clean |
| UVs | smart_uv_unwrap recipe, then a checker render to confirm |
| Textures | Wood, iron and gold images, with roughness and tileable normal maps derived in Python, then pbr_material_from_textures |
| Look | three_point_lighting and set_render_preset, EEVEE previews and a Cycles hero render on the GPU (about 9 s) |
| Review | A BVH interference check found coins clipping through the lid, and fixed them from the numbers, not screenshots |
| Budget | xr_budget_report (Quest): 32,304 tris (4.3%), 102 draw calls (68%), 5.3 MB textures. All within budget |
| Export | export_asset(preset="webxr_glb"): an 8.4 MB Draco-compressed GLB, plus the .blend, a render of each stage, a 60-frame turntable and a caption per stage |
Next steps
- Merge the 70 coins into one mesh to cut draw calls before shipping to Quest.
- Run
generate_lodsif the chest will be seen from a distance.
Connects securely to your tools. The creator never sees your data.
About this skill
What it does
BlendPro teaches your agent to work in Blender the way an experienced technical artist does, and to do it cheaply. Most agent-driven Blender work runs code, takes a screenshot to see what happened, and repeats. Images are the most expensive thing a model reads, and a screenshot can't show topology, collisions or performance budgets anyway.
This skill gives the agent a text-first loop instead:
- Read the scene as text: one compact line per object.
- Edit in one call: batched edits, whole node trees as JSON, or one of 100 tested recipes.
- Read the diff that every edit returns, so there's no re-query.
- Verify with text: mesh topology reports, spatial checks (what rests on what, real intersections, camera framing) and API lookups against the installed Blender version.
- Render once at the end to judge the look.
What's inside
- Workflow guide for the BlendPro MCP tools: when to use
run_python,node_tree_build,mesh_analysis,spatial_report, recipes, checkpoints and export presets. - Curated Blender know-how: character blockouts, boolean cleanup and quad retopology, beginner stylized sculpting, texture painting, Thin Wall transparency, stylized fur and feathers in Geometry Nodes, stylized EEVEE, and recommended free add-ons with exact menus and hotkeys.
- Blender internals reference: DNA/RNA, the depsgraph and evaluated copies, mesh attributes and bmesh, operator context, GN fields, layered actions and 5.x API changes, so the agent writes bpy that works the first time.
- Notes verified against Blender 5.2.2.
Requirements
The know-how works on its own for questions and scripts. To let the agent drive Blender directly, install the free, open-source BlendPro MCP server (MIT): https://github.com/MemeGamer/blendpro-mcp. You need Blender 4.2+ (tested on 5.2.2), Python 3.10+ and uv.
Proof
The screenshots show a textured treasure chest built end to end through BlendPro in a headless Blender: blockout, cleanup, UVs, ChatGPT-generated PBR textures, EEVEE and Cycles renders, a Quest budget check and a Draco GLB for WebXR. The GitHub README lists every stage and the measured token cost.
MediaPipe requires humanoid figures facing the -Y axis. Voxel remesh values assume 1 to 2 meter objects with scale applied. MCP requires local port 9877.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 10 days ago
- Free with an Agensi account
Requires