Origin MCP
Enables AI clients to drive a running OriginLab Origin instance, allowing LabTalk scripting, worksheet read/write, and project management.
README
Origin MCP
An OriginLab Origin app that lets AI clients drive your running Origin instance.
Install the app, click it once, and Origin starts a small background server (an MCP server). Point any MCP-capable AI client at it and the AI can now work inside this Origin: run LabTalk, read and write worksheets, run fits, create projects, and read results back.
This is a standalone app — it is not part of Batalyse.
What the AI can do
Once connected, the client has tools to:
- Run LabTalk — any script, with or without reading values back (
run_labtalk,run_labtalk_with_readback) - Read/write worksheet data (
get_worksheet_data,set_worksheet_data,add_worksheet) - Read/write LabTalk variables and the project tree (
get_labtalk_value,set_labtalk_var,get_labtalk_tree) - Manage projects (
create_project,save_project)
Quick start
- Install the
OriginMCP.opx(double-click it, or drag it onto Origin). On install, Origin downloads the server's Python dependencies in the background — the first install takes a minute. - Click the "Origin MCP" app in the Apps Gallery to start the server. A
message box confirms it's running. Click again to stop. The server listens on
http://127.0.0.1:8000/mcp. - Point your MCP client at the URL (see below) and have it call
connect_origin.
That's it — the server runs in the background while Origin is open and shuts itself down when Origin closes.
Connect your AI client
The app starts the server for you, so the client just connects to the URL:
{
"mcpServers": {
"origin-mcp": { "type": "http", "url": "http://localhost:8000/mcp" }
}
}
Auto-start at Origin launch (optional)
Clicking the app starts the server. To start it automatically every time Origin launches, add this to your Origin startup script:
run.section("%@A%@X\launch.ogs", autostart);
It's safe to call repeatedly — it does nothing if the server is already running.
Connecting from WSL
A WSL2 client can't reach Windows localhost, so the app binds the server in a
way WSL can reach (it is not exposed to your LAN — Windows Firewall still
blocks inbound on your physical network adapters).
- Connect to
http://host.docker.internal:8000/mcp(transport:http). - Allow the WSL subnet to reach the port. In an elevated PowerShell:
(New-NetFirewallRule -DisplayName "Origin MCP (WSL->host 8000)" -Direction Inbound ` -Action Allow -Protocol TCP -LocalPort 8000 -RemoteAddress 172.16.0.0/12172.16.0.0/12is the private WSL/Hyper-V range — not your Wi-Fi/LAN.)
⚠️ The server runs arbitrary LabTalk, so keep the firewall scope tight.
How it works
Clicking the app runs a small lifecycle manager inside Origin's embedded Python, which launches a separate, hidden sidecar process for the actual server. The sidecar attaches back to this Origin over COM and serves the MCP tools:
Origin (this instance)
│ click "Origin MCP" → launch.ogs → manage.py (start / stop / toggle)
▼
manage.py ── launches (hidden) ──► sidecar: mcp_bootstrap.py → origin_mcp_server.py
│ records host PID + a token │ attaches back to THIS Origin (COM)
▼ ▼
AI client ──► http://127.0.0.1:8000/mcp
The sidecar runs off Origin's UI thread, so the server never freezes Origin, and it exits on its own when Origin closes — no orphaned processes left behind.
Implementation details (the three-process design, COM lifecycle, and packaging)
live in CLAUDE.md.
Files
| File | Role |
|---|---|
launch.ogs |
App entry: click = toggle; [start]/[stop]/[status]/[setup]/[autostart] |
manage.py |
Lifecycle manager (runs in Origin's embedded Python) |
mcp_bootstrap.py |
Sidecar entry: attaches to the right Origin, then runs the server |
origin_mcp_server.py |
The MCP server and its tools |
AfterInstall.ogs |
Installs the server's Python dependencies on install |
BeforeUninstall.ogs |
Stops the sidecar before the app is removed |
Notes
AppIcon.pngis a placeholder — replace with Origin MCP branding before public distribution.- Single instance: the OS port
:8000acts as the lock. A second Origin instance that tries to start a server will fail to bind and exit quietly; the first instance keeps serving.
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。