Origin Pro MCP Skill / Origin Pro MCP 技能

English

Use this skill when the user wants to create, edit, style, fit, or export Origin Pro worksheets and figures through the origin-pro MCP server.

Scope

• Create and manage Origin projects.
• Create worksheets, write JSON column arrays, read worksheet data, and import numeric CSV files.
• Create scatter, line, line+symbol, bar, histogram, box, contour, pie, bubble, and selected 3D graph types.
• Apply publication-style formatting with one tool call.
• Run linear and nonlinear curve fitting.
• Execute LabTalk for advanced Origin operations.
• Export verified image files from Origin graph pages.

Prerequisites

• Windows with Origin Pro 2020 or newer.
• Windows Python 3.10 or newer.
• Package installed with python -m pip install -e ..
• MCP server configured as origin-pro.
• Origin Pro started before COM-dependent operations.

Default Workflow

1. Confirm Origin Pro is running.
2. Use new_project only when it is acceptable to clear the current Origin session.
3. Load data with create_worksheet + set_worksheet_data or import_csv_to_worksheet.
4. Use create_graph and add_plot_to_graph.
5. Use apply_publication_style for journal-ready formatting.
6. Use curve_fit when fitting is requested.
7. Use export_graph or export_all_graphs, then verify the returned file path and size.

Tool List

Project:

• new_project
• save_project
• load_project

Worksheet and data:

• create_worksheet
• set_worksheet_data
• get_worksheet_data
• import_csv_to_worksheet
• list_worksheets

Graphing:

• create_graph
• add_plot_to_graph
• set_axis_labels
• set_axis_range
• export_graph
• export_all_graphs

Styling:

• apply_publication_style
• set_plot_style
• set_graph_font
• set_legend
• set_tick_style

Analysis:

• curve_fit
• list_fitting_functions

Advanced:

• run_labtalk
• get_labtalk_variable

Safety Rules

• Do not include local API keys, token files, Zotero settings, or workspace-specific MCP config in shared outputs.
• Ask before calling new_project if the user may have unsaved Origin work.
• Treat a successful export as valid only when the tool reports a generated file with non-trivial size.
• If a tool hangs, check whether Origin is showing a modal dialog.

中文

当用户希望通过 origin-pro MCP server 创建、编辑、排版、拟合或导出 Origin Pro 工作表和图像时，使用本技能。

能力范围

• 创建和管理 Origin 项目。
• 创建工作表、写入 JSON 列数组、读取工作表数据、导入数值型 CSV 文件。
• 创建 scatter、line、line+symbol、bar、histogram、box、contour、pie、bubble 和部分 3D 图。
• 一键应用论文图样式。
• 执行线性和非线性曲线拟合。
• 通过 LabTalk 执行高级 Origin 操作。
• 从 Origin 图页导出经过文件校验的图片。

前置条件

• Windows，安装 Origin Pro 2020 或更高版本。
• Windows Python 3.10 或更高版本。
• 已执行 python -m pip install -e .。
• MCP server 已配置为 origin-pro。
• 依赖 COM 的操作前请先启动 Origin Pro。

默认流程

1. 确认 Origin Pro 正在运行。
2. 只有在可以清空当前 Origin 会话时才调用 new_project。
3. 使用 create_worksheet + set_worksheet_data 或 import_csv_to_worksheet 加载数据。
4. 使用 create_graph 和 add_plot_to_graph 绘图。
5. 使用 apply_publication_style 完成论文图样式。
6. 需要拟合时调用 curve_fit。
7. 使用 export_graph 或 export_all_graphs 导出，并检查返回的路径和文件大小。

工具清单

项目管理：

• new_project
• save_project
• load_project

数据与工作表：

• create_worksheet
• set_worksheet_data
• get_worksheet_data
• import_csv_to_worksheet
• list_worksheets

绘图：

• create_graph
• add_plot_to_graph
• set_axis_labels
• set_axis_range
• export_graph
• export_all_graphs

样式：

• apply_publication_style
• set_plot_style
• set_graph_font
• set_legend
• set_tick_style

分析：

• curve_fit
• list_fitting_functions

高级：

• run_labtalk
• get_labtalk_variable

安全规则

• 不要在共享输出中包含本地 API key、token 文件、Zotero 设置或工作区专用 MCP 配置。
• 如果用户可能有未保存的 Origin 工作，调用 new_project 前先确认。
• 只有当导出工具报告生成了非空文件时，才视为导出成功。
• 如果工具卡住，先检查 Origin 是否弹出了模态对话框。

Origin Pro MCP Skill Package

English | 中文说明

Origin Pro MCP Skill Package is a Windows-first MCP server and Codex/Claude-style skill package for controlling OriginLab Origin Pro through COM automation. It lets an AI assistant create workbooks, write and read worksheet data, import CSV files, build graphs, apply publication-style formatting, run curve fitting, execute LabTalk, export figures, and save/load Origin projects.

This repository is a sanitized, reusable package. It does not contain personal API keys, Zotero configuration, local workspace state, generated experiment outputs, or machine-specific MCP configuration.

Features

• Origin Pro COM automation through Origin.ApplicationSI
• 23 MCP tools for project, worksheet, graph, style, fitting, export, and LabTalk workflows
• File-verified graph export using Origin expGraph
• CSV import that respects the requested Origin workbook name and avoids Origin impasc side effects
• Publication figure workflow in skills/publication-figure.md
• Pytest coverage with Origin-aware skip behavior for machines without Origin installed

Requirements

• Windows
• Origin Pro 2020 or newer, installed and able to start
• Python 3.10 or newer, using Windows Python rather than WSL Python
• Python packages: mcp, pywin32, Pillow

Install

git clone https://github.com/leima-max/origin-pro-mcp-skill.git
cd origin-pro-mcp-skill
python -m pip install -e .

Start Origin Pro before calling tools that need COM automation.

MCP Configuration

Use either the installed console command:

{
  "mcpServers": {
    "origin-pro": {
      "command": "origin-pro-mcp"
    }
  }
}

Or run the local server file directly:

{
  "mcpServers": {
    "origin-pro": {
      "command": "python",
      "args": ["-u", "PATH_TO_REPO/server.py"]
    }
  }
}

A sanitized example is available at examples/mcporter.example.json.

Quick Smoke Test

python -m pytest -q

On machines without Origin Pro, Origin-dependent tests are skipped. On a Windows machine with Origin Pro available, the integration tests exercise the real COM workflow.

You can also verify tool registration without starting Origin:

python -c "import origin_pro_mcp.server; from origin_pro_mcp.app import mcp; print(len(mcp._tool_manager._tools))"

Expected output: 23.

Typical AI Workflow

1. Start Origin Pro.
2. Ask the assistant to create or import worksheet data.
3. Create a graph from worksheet columns.
4. Apply publication style in one call.
5. Fit a curve if needed.
6. Export figures and verify that files were created.

Example tool flow:

new_project()
create_worksheet(book_name="Data", sheet_name="Sheet1")
set_worksheet_data(
    book_name="Data",
    sheet_name="Sheet1",
    columns="[[0,1,2,3],[0.02,0.08,0.18,0.36],[0.11,0.42,0.90,1.62]]",
    column_names="Bias,Dark,Light"
)
create_graph(graph_name="Fig1", data_book="Data", data_sheet="Sheet1", x_col=1, y_col=2, plot_type="line+symbol")
add_plot_to_graph(graph_name="Fig1", data_book="Data", data_sheet="Sheet1", x_col=1, y_col=3, plot_type="line+symbol")
apply_publication_style(
    graph_name="Fig1",
    x_label="Voltage (V)",
    y_label="Current density (mA cm^-2)",
    x_min=0,
    x_max=3,
    y_min=0,
    y_max=2,
    legend_entries="Dark,Light",
    legend_position="top-left"
)
export_graph(graph_name="Fig1", file_path="OUTPUT_DIR/Fig1.png")

MCP Tools

Project:

• new_project
• save_project
• load_project

Worksheet and data:

• create_worksheet
• set_worksheet_data
• get_worksheet_data
• import_csv_to_worksheet
• list_worksheets

Graphing:

• create_graph
• add_plot_to_graph
• set_axis_labels
• set_axis_range
• export_graph
• export_all_graphs

Styling:

• apply_publication_style
• set_plot_style
• set_graph_font
• set_legend
• set_tick_style

Analysis:

• curve_fit
• list_fitting_functions

Advanced:

• run_labtalk
• get_labtalk_variable

Notes and Known Constraints

• Origin COM only works from Windows Python.
• The first COM connection can take several seconds if Origin is cold-starting.
• If Origin shows a blocking dialog, MCP calls may wait or time out.
• export_graph uses expGraph with a directory path and a separate filename; the tool verifies output file existence and size.
• import_csv_to_worksheet parses CSV in Python and writes columns through COM to avoid impasc auto-generated sparkline graph pages.

Security and Sanitization

This package intentionally excludes:

• local config/mcporter.json
• API keys and token files
• Zotero or other third-party connector settings
• local workspace memory, logs, and generated experiment output
• user-specific absolute paths

Before publishing derived versions, run a secret scan for keys, tokens, passwords, local connector names, and machine-specific paths. For example:

rg -n "key|token|secret|credential|password|local_user_path|connector_specific_env"

License

MIT. The original upstream package is credited in pyproject.toml project metadata and source history notes.