Skip to content

CLI Reference

The mcp-ai-agent-guidelines CLI provides commands for project onboarding, session lifecycle management, memory inspection, and observability reporting.

Terminal window
# Run via npx (no install required)
npx mcp-ai-agent-guidelines [command]
# Or after local install
node dist/cli-main.js [command]

OptionDescription
-v, --verboseEnable verbose output
--versionPrint version
--helpShow help

Run the interactive onboarding wizard. Guides you through model configuration, skill surface selection, and MCP client setup. Skips automatically if configuration already exists.

Terminal window
mcp-ai-agent-guidelines onboard init

Show the current onboarding status and active configuration.

Terminal window
mcp-ai-agent-guidelines onboard status

Reset configuration. Removes session and memory state and re-runs the wizard.

Terminal window
mcp-ai-agent-guidelines onboard reset

Manage model routing via orchestration.toml without editing the file directly.

Print the current orchestration.toml as JSON.

Terminal window
mcp-ai-agent-guidelines orchestration config read

Update a specific role in orchestration.toml.

Terminal window
mcp-ai-agent-guidelines orchestration config write --role strong_primary --model claude-sonnet-4-5

List all memory artifacts and TOON snapshots.

Terminal window
mcp-ai-agent-guidelines memory list

Print a specific memory artifact by ID.

Terminal window
mcp-ai-agent-guidelines memory show <session-id>

Print version, loaded skills, active flags, and model router state.

Terminal window
mcp-ai-agent-guidelines status

These npm scripts delegate to the docs/ Starlight site.

Terminal window
npm run docs:dev # Start dev server at localhost:4321
npm run docs:build # Build static site to docs/dist/
npm run docs:preview # Preview built site locally

The MCP server itself (not the CLI) is started with:

Terminal window
node dist/index.js

This speaks MCP over stdin/stdout and is what your MCP client (Copilot, Claude Desktop, etc.) invokes.