blender-mcp-synthgen
MCP server for procedural 3D synthetic-data generation in Blender, providing 33 grounded tools for node graph authoring, scene-graph traversal, and parameter sweeps with schema verification.
README
blender-synthgen-mcp
Procedural 3D synthetic-data toolkit for Blender — grounded node schemas, a lazy scene-graph walker, and a Houdini→Blender knowledge base, delivered as a self-contained Blender addon with an MCP server built in.
Quick Install (Blender addon)
-
Download the latest
synthgen_mcp.zipfrom releases (or build it yourself — see below) -
Install in Blender: Edit → Preferences → Add-ons → Install from Disk → select the zip
-
Enable "Synthgen MCP" — first enable installs dependencies (~15s one-time)
-
Connect from your IDE:
Claude Code / Cursor / VS Code — add to your MCP config:
{"synthgen": {"url": "http://localhost:8400/sse"}}
That's it. The addon runs an SSE MCP server inside Blender on port 8400. Your IDE connects to it and gets 33 grounded tools for procedural 3D work.
N-panel controls
In the 3D viewport sidebar (N) → "Synthgen MCP" tab:
- Start/Stop — toggle the MCP server
- Copy MCP Config — copies the connection JSON to your clipboard
- Server status — shows running state and port
Change the port
Edit → Preferences → Add-ons → Synthgen MCP → Port (default 8400)
What's inside
The addon bundles:
- 33 MCP tools across 5 layers (setup, procedural authoring, graph introspection, pipeline, verification)
- Grounding data — extracted node schemas for Blender 5.2 (Geometry Nodes, Shader, Compositor)
- Main-thread executor — marshals all bpy calls safely from the MCP background thread
Tool layers
| Layer | Tools | Purpose |
|---|---|---|
| Setup | create_object, create_material, assign_material, set_parent, configure_render, import_asset, edit_mesh, add_keyframes |
Scene building blocks |
| Procedural | add_gn_modifier, add_node, link_sockets, set_node_property, set_socket_default, expose_parameter, add_driver, wire_attr_bridge, wire_compositor_pass |
Node graph authoring with grounded identifiers |
| Introspection | graph_nodes, graph_neighbors, graph_reachable, graph_impact_set, graph_attribute_trace, graph_snapshot |
Live scene-graph traversal |
| Pipeline | set_parameter, render, sweep, export_labels |
Parameter sweeps + provenance |
| Schema | schema_find, schema_show, schema_socket, schema_setting |
Query node schemas |
| Verify | verify_attribute_exists |
Runtime attribute checks |
| Escape hatch | execute_python |
Direct bpy access (ungrounded) |
Why this exists
Transport (a Blender MCP) is a commodity. The moat is version-exact grounding + a build→verify loop. LLMs hallucinate node/socket identifiers, so we extract the real schema from Blender itself and give the agent tools to query it — never trusting model memory. The same idea, one level up, becomes the scene-graph walker: a lazy, pull-based view over the live scene so the agent can traverse real relationships.
Governing principle (see knowledge/procedural_paradigm.md):
derive, don't set. Native Blender is imperative/destructive; the agent must live on the
procedural island (Geometry Nodes / shader / compositor / drivers), because you cannot sweep
a hand-edit, and sweeping a parameter space is the whole job.
Build from source
# Install dev dependencies
pip install -e ".[dev]"
# Run tests (offline — no Blender needed)
pytest
# Build the addon zip
python scripts/build_addon.py
# → dist/synthgen_mcp.zip
Development mode
For development without rebuilding the zip each time:
- Set Blender's
BLENDER_USER_SCRIPTSenvironment variable to point to theaddon/directory - Or symlink
addon/synthgen_mcp/into Blender's addons folder
The addon auto-detects whether it's running from source (adds src/ to path) or from a
built zip (uses bundled files).
Grounded against
- Blender 5.2.0 LTS — Geometry Nodes (360), Shader (118), Compositor (168) node schemas
- Houdini 22.0.368 — SOP/VOP/COP node schemas (for the crosswalks)
Regenerate for another version with the extractors in src/synthgen/extract/ (run inside
Blender / hython); drop results in data/schemas/<app-version>/.
Schema CLI (standalone)
The schema query tool works without Blender:
python -m synthgen.schema.query stats
python -m synthgen.schema.query show "Store Named Attribute"
python -m synthgen.schema.query --shader show "Attribute"
python -m synthgen.schema.query --compositor find "Cryptomatte"
Architecture
IDE (Claude Code / Cursor / VS Code)
│
│ MCP protocol over SSE HTTP
│
▼
Blender addon (synthgen_mcp)
├── SSE MCP server (background thread, port 8400)
├── Main-thread executor (bpy.app.timers queue)
├── 33 grounded tools
├── Bundled schema data
└── N-panel UI
See ROADMAP.md for project status and phase details.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.