Editors and agents
npx -y @sylphx/lockdocs setup writes these for you. To do it by hand:
| Client | Config | Entry |
|---|---|---|
| Claude Code | claude mcp add --scope user lockdocs -- npx -y @sylphx/lockdocs mcp | |
| Codex | ~/.codex/config.toml | [mcp_servers.lockdocs] command = "npx", args = ["-y", "@sylphx/lockdocs", "mcp"] |
| Cursor | ~/.cursor/mcp.json | mcpServers.lockdocs |
| VS Code | <config>/Code/User/mcp.json | servers.lockdocs with "type": "stdio" |
| Claude Desktop | <config>/Claude/claude_desktop_config.json | mcpServers.lockdocs |
| Windsurf | ~/.codeium/windsurf/mcp_config.json | mcpServers.lockdocs |
| Gemini CLI | ~/.gemini/settings.json | mcpServers.lockdocs |
<config> is ~/.config on Linux, ~/Library/Application Support on macOS and %APPDATA% on Windows. On Windows the command is cmd /c npx -y @sylphx/lockdocs mcp.
Environment
| Variable | Effect |
|---|---|
LOCKDOCS_ROOT | Project directory when the client sends no roots |
LOCKDOCS_FETCH=1 | Allow fetching exact versions that are not installed |
LOCKDOCS_CACHE | Cache directory (default: the OS cache dir + /lockdocs) |
LOCKDOCS_NO_SYSTEM_PYTHON=1 | Do not ask the system interpreter for its site-packages |