Serpentine3D MCP Server
A Serpentine3D MCP server lets an MCP client inspect and drive a live NURBS modeling session.
Inspect the scene (objects, layers, selection, bounds, display mode) and capture PNG screenshots.
Create NURBS curves and surfaces/solids (extrude, revolve, loft, planar, sweep).
Perform booleans, transforms, and selection by name/kind/layer.
Run any Serpentine3D command as if typed, with interactive inputs.
Prepare editable Python script drafts without executing them.
Manage layers, visibility, colors, parent layers, and drawing styles.
Import/export .serp, .step/.stp, and .obj files.
Measure distance, length, area, volume, bbox, and centroid.
Undo/redo operations and adjust viewport/view/zoom/display mode.
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., "@Serpentine3D MCP Server loft a surface between two curves"
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.
Serpentine3D
The open-source NURBS modeller for things that get built. Exact curves and surfaces, drawing sheets a workshop can build from, a command line that speaks Rhino, and an assistant that can drive all of it. Linux, Windows and macOS. Free, with no account.
Download · Website · Documentation · Coming from Rhino?
https://github.com/user-attachments/assets/837002ba-7780-49d3-b966-71d6469ba884
Made in Serpentine3D
Related MCP server: Rhino MCP
What it does
Exact geometry. NURBS curves, surfaces and solids on the OpenCASCADE kernel, not meshes. Rhino
.3dm, STEP, DXF, SVG, OBJ, FBX and STL in and out; glTF, USD and 3MF out; E57 point clouds in.A command line that speaks Rhino.
loft,sweep2,filletedge,booleanunionand more than 200 others, with Rhino's aliases, object snaps, layers and a gumball that pushes faces and rounds edges by hand.Drawings from the model. A4 to A0 sheets with live detail views, hidden-line views, dimensions, hatching and title blocks, printed to vector PDF.
An assistant that can see. Describe a change and it runs the commands, looks at the viewport to check its work and fixes its own mistakes, all undoable. Use Claude, a ChatGPT account, an OpenAI key, or a local, open-weight model.
Headless. Script it in the built-in editor, batch it with
serp3d-batch, or let any MCP client drive a live session.Checks before you build. Zebra stripes, curvature, draft analysis and
printcheckfor 3D printing.
Download
Platform | Download | Size | Notes |
Linux | 509 MB | Make it executable and run it. Nothing to install. | |
Windows | 119 MB | Unsigned for now: choose More info, then Run anyway. | |
macOS | 268 MB | Apple Silicon, macOS 14 or newer. Unsigned for now: right-click the app and choose Open. |
Each build carries the OpenCASCADE kernel and its own Python, and needs a GPU with OpenGL 3.3. Headless use works anywhere. Release notes · Code signing policy
Script it
# make_part.py, run with: serp3d-batch make_part.py crate.step
box = doc.add(geo.make_box((0, 0, 0), 100, 100, 100), name="Crate")
doc.run("filletedge", ["Crate", "", "5"])
doc.export(args[0] if args else "crate.step")doc.run drives any command by answering its prompts. More in the
scripting guide and the
assistant and MCP guide.
Documentation
Start here | How-to | Reference |
Build from source
git clone https://github.com/chisomobanzi/Serpentine3D && cd Serpentine3D
python3 -m venv .venv && .venv/bin/pip install -e ".[dev]"
.venv/bin/serp3d # the app
.venv/bin/pytest # the testsPython 3.10 or later. The kernel installs as pip wheels (cadquery-ocp), with
no conda and no system packages, on Linux, Windows and macOS alike.
Support
Serpentine3D is free and always will be: no subscription, no licence server, no "upgrade to Pro". One person builds it between other work. If it is useful to you, a small contribution on Ko-fi keeps it moving.
Bug reports, sample files and documentation fixes are worth as much and cost nothing: see CONTRIBUTING.md, and Not here yet for what is missing.
Named after the serpentine stone of Zimbabwean Shona sculpture, and the S-curve at the heart of NURBS geometry. MIT licensed.
Available Tools
15 toolsserp_booleanA
Boolean operation between solids: 'union', 'difference' (targets minus tools), or 'intersection'. Tools are consumed.
Inputs: tools (required), targets (required), operation (required). Returns: result
serp_commandA
Run any Serpentine3D command exactly as if typed in the command line, supplying its interactive inputs in order.
Examples: command='circle', inputs=['0,0,0', '5'] command='extrude', inputs=['Curve 01', '', '20'] (object name(s), then '' to end selection, then distance) command='zoomextents'
Selection prompts accept object names, 'all', or '' to finish. Too few inputs is an error, and so is one left over when the command finishes: the error says which prompt each input answered.
Inputs: command (required), inputs. Returns: result
serp_create_curveB
Create a NURBS curve from 3D points [[x,y,z], ...].
kind: 'interp' (curve passes through the points), 'control' (points are control vertices), 'polyline' (straight segments), 'line' (two points).
Inputs: points (required), kind, name, closed, +1 more. Returns: result
serp_create_surfaceA
Create a surface/solid from existing curves (referenced by name).
operation: 'extrude' params: direction [x,y,z] (default [0,0,1]), distance (default 10), cap (bool, default true - closed profiles become solids) 'revolve' params: axis_point, axis_dir, angle (degrees, default 360) 'loft' curves: 2+ profiles in order; params: ruled (bool) 'planar' flat surface from one closed planar curve 'sweep' curves: [profile, rail]
Inputs: curves (required), operation (required), name, params. Returns: result
serp_exportA
Export the scene (or current selection) to .serp, .step/.stp or .obj — format chosen by extension.
Inputs: path (required), selected_only. Returns: result
serp_importA
Add a file to the scene (.serp, .step/.stp, .obj). What is already there stays. The user's view is left alone unless zoom_extents is true.
Inputs: path (required), zoom_extents. Returns: result
serp_layersA
Manage layers. action: 'list', 'create', 'rename', 'visible', 'current', 'color' (color as [r,g,b] 0-1), 'assign' (move objects to layer), 'delete', 'style' (how a layer draws: style={linetype, print_width mm (0 = printer default), screen_width px or 'from print', hatch none/lines/cross/solid or a pattern loaded with the hatchpatterns command, hatch_angle degrees, hatch_spacing mm on paper, hatch_color ink/layer/#rrggbb, hatch_background none/layer/#rrggbb (what a cut shows under its hatch), cut_line_width mm on paper round a cut (0 = none), cut_line_color ink/layer/#rrggbb}). A layer can live under another: pass parent to 'create', and name a layer by its path ('Walls::Interior') where two layers share a name.
Inputs: name, color, style, action, +4 more. Returns: result
serp_measureA
Measure geometry. what: 'distance' (needs points=[[..],[..]]), 'length', 'area', 'volume', 'bbox', 'centroid' (need targets).
Inputs: what (required), points, targets. Returns: result
serp_prepare_scriptA
Open Python in a new editable Script draft without executing it.
Existing drafts and geometry are preserved. Scripts can use doc (a working document), geo (geometry helpers), and selected (selected SceneObjects). The user can edit the draft, Run to preview, then Keep or Discard its changes.
Inputs: source (required), title. Returns: result
serp_scene_infoA
Get the current scene: objects (name/kind/layer/bbox, and the user_text keys and values of any that carry some), layers, selection, bounds, and display mode. Set user text with the setusertext command.
Inputs: no named parameters. Returns: result
serp_screenshotA
Capture the current 3D viewport so you can see the model. Returns a PNG image. Use after making changes to check the result.
Inputs: width.
serp_selectA
Select objects by names, kind (curve/surface/solid), or layer name. mode: 'replace', 'add', or 'clear' (clear ignores other args).
Inputs: kind, mode, layer, names. Returns: result
serp_transformA
Transform objects (by name).
operation/params: 'move' offset [dx,dy,dz] 'copy' offset [dx,dy,dz] 'rotate' center [x,y,z], axis [x,y,z] (default Z), angle degrees 'scale' center, factor (uniform) or factors [sx,sy,sz] 'mirror' plane_point, plane_normal, keep_original (bool)
Inputs: targets (required), operation (required), params. Returns: result
serp_undoA
Undo the last operation (or redo with redo=true).
Inputs: redo. Returns: result
serp_viewportA
Adjust the viewport. view: top/front/right/left/back/bottom/ isometric/perspective. display_mode: wireframe/shaded/ghosted. zoom_extents fits all objects in view.
Inputs: view, display_mode, zoom_extents. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 15 tools
Most specialized tools have clearly distinct purposes, but serp_command is a universal escape hatch that can duplicate almost every other tool. serp_layers also bundles many actions under one name, and agents may hesitate between using a direct tool and the generic command tool.
All names use the same serp_ snake_case prefix, which is highly consistent. However, they are not strictly verb_noun: several are nouns or broad concepts (viewport, layers, scene_info, screenshot, command).
With 15 tools, the set sits comfortably in the well-scoped range for a 3D CAD server. Each tool maps to a major workflow area such as creation, transformation, selection, measurement, or export.
The surface covers most of the modeling lifecycle: curves, surfaces, booleans, transforms, layers, measurement, import/export, undo, viewport, and scripting. Explicit delete/remove and primitive-creation operations are missing, though serp_command and serp_prepare_script provide workarounds.
Maintenance
Related MCP Connectors
Parametric 3D CAD for AI agents: build print-ready parts, check them, export STL, 3MF or STEP.
A browser 3D CAD for your AI: it scripts real-size models, uCAD builds them and returns pictures.
CAD in ChatGPT and Claude: editable parts, STEP/STL/drawings, in-chat 3D, automatic checks.
Generate 3D models, worlds, PBR materials, and coordinated asset packs from AI agents.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables AI-assisted 3D modeling and scene control in Autodesk Maya through natural language commands, supporting object creation, transformation, material application, and more.63 npm8MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI-powered 3D modeling in Rhino 8 through 135+ tools for geometry, transformations, booleans, layers, materials, and more, using natural language.7MIT
- AlicenseAqualityDmaintenanceEnables AI-driven 3D modeling in Blender by providing tools to create primitives, apply modifiers and materials, set up lighting and cameras, capture viewport snapshots, export assets, inspect scenes, and execute Python commands via natural language.10MIT
- AlicenseCqualityDmaintenanceEnables AI-driven 3D modeling, analysis, diagnostics, and Grasshopper integration within Rhino 3D through 90+ tools via the Model Context Protocol.1053MIT