cura-mcp
Enables AI agents to slice 3D models (STL files) headlessly using UltiMaker Cura's CuraEngine, returning ready-to-print G-code and estimated print time.
README
cura-mcp
An MCP server that lets an AI agent (Claude, etc.) slice 3D models with UltiMaker Cura
(the CuraEngine backend) headlessly — no GUI, no clicking. Load STL files, pick a printer
profile, override any Cura setting, and get back a ready-to-print .gcode plus the estimated
print time.
Built because Cura has no official MCP and its engine is awkward to run standalone — this wraps it cleanly.
What it exposes
| Tool | What it does |
|---|---|
list_printers |
List available printer profiles (id, build volume, nozzle). |
get_profile |
Return the full settings of a profile, to inspect or tweak. |
slice |
Slice one or more STL files with a profile + optional setting overrides → gcode + print time. |
Example: "Slice part.stl on the Creasee 340 with 0.16 mm layers, 5 walls, 30% infill, and a raft."
→ the agent calls slice(stl_paths=["/…/part.stl"], printer_id="creasee340", settings={"layer_height":0.16,"wall_line_count":5,"infill_sparse_density":30,"adhesion_type":"raft"}).
Why this was hard (and how it's solved)
CuraEngine from the Cura AppImage doesn't run standalone: it needs (1) the AppImage's bundled
ld-linux interpreter, (2) all of the AppImage's shared libraries, and (3) a complete, flattened
set of settings (fdmprinter + fdmextruder defaults + your overrides) — miss one and it aborts.
engine.py handles all three: it runs CuraEngine via the bundled linker with a full library path,
and flattens Cura's nested settings tree into the ~400 -s key=value flags CuraEngine expects.
Requirements
- Linux, Python ≥ 3.10.
- UltiMaker Cura AppImage extracted to
~/opt/cura/squashfs-root(or setCURA_ROOT):cd ~/opt/cura && /path/to/UltiMaker-Cura-*.AppImage --appimage-extract
Install
git clone <this repo> && cd cura-mcp
python3 -m venv .venv && ./.venv/bin/pip install -e .
Use with Claude Code
claude mcp add cura-slicer -- /ABS/PATH/cura-mcp/.venv/bin/python -m cura_mcp
Then restart your session and ask Claude to slice a model.
Printer profiles
Profiles live in src/cura_mcp/profiles/*.json as {name, manufacturer, description, settings:{…Cura settings…}}.
Ships with Creasee 340 (large CR-10-class bowden printer, CR-Touch ABL, quality profile tuned for a
slightly warped bed) and a generic 220 mm printer. Add your own by dropping a JSON file in that folder.
Notes
- Print time and filament are computed from the actual toolpaths (analyze_gcode), so they vary with settings and model — reliable estimates.
- License AGPL-3.0 (CuraEngine is AGPL; this wrapper follows).
Author
Built by Claude (Anthropic's Opus 4.8), autonomously, for Michael Gagnon — solving the standalone-CuraEngine problem, the settings flattener, the MCP server, and a real gcode-based time/filament estimator. Tested with a 9-case suite (tools, overrides, multi-part, error handling, stat-variation).
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。