ScopeOne Skill

Agent operating guidance and a Claude Code plugin for controlling a running ScopeOne microscopy application through its bundled MCP server.

This repository does not contain ScopeOne itself. Install ScopeOne first and locate ScopeOneMcpServer.exe beside ScopeOne.exe. On Linux, use the corresponding executable without the .exe suffix.

Claude Code

Add this repository as a plugin marketplace and install the ScopeOne plugin:

/plugin marketplace add Experimental-Microscopy-Lab/ScopeOne-Skill
/plugin install scopeone@scopeone
/reload-plugins

When prompted, select the ScopeOneMcpServer executable from the ScopeOne installation. Start ScopeOne before asking Claude to inspect or control it. The plugin supplies both the MCP configuration and the ScopeOne operating Skill.

Codex

Ask Codex to install the Skill directly from GitHub:

Install the ScopeOne Skill from
https://github.com/Experimental-Microscopy-Lab/ScopeOne-Skill/tree/main/skills/scopeone

Codex installs it through its built-in Skill Installer. The Skill explains how to locate ScopeOneMcpServer and prepares the exact MCP registration command. Because MCP tools are loaded when Codex starts, finish setup outside the active session:

1. Close Codex.
2. Run the prepared codex mcp add scopeone -- "ABSOLUTE_SERVER_PATH" command in a normal terminal.
3. Start ScopeOne and reopen Codex.
4. Ask Codex to use the ScopeOne Skill to inspect or control ScopeOne.

MCP registration is a one-time setup. Do not launch ScopeOneMcpServer manually; Codex starts it when the registered tools are needed.

For an offline installation, copy or link skills/scopeone to $CODEX_HOME/skills/scopeone, or to ~/.codex/skills/scopeone when CODEX_HOME is unset.

Contents

• skills/scopeone: shared ScopeOne Skill for Claude Code and Codex
• .mcp.json: Claude Code MCP server configuration
• .claude-plugin/plugin.json: Claude Code plugin metadata
• .claude-plugin/marketplace.json: Claude Code marketplace metadata

ScopeOne remains the authority for operation schemas, validation, safety classifications, and hardware read-back. Review physical, destructive, and file-writing operations before allowing an agent to execute them.