Using Otto

Once the Otto MCP server is connected to your AI coding assistant — Claude Code, GitHub Copilot, or Cursor — you can interact with Otto through VS Code or the CLI. This guide covers both approaches.

Activating Otto

Otto only responds when you address it by name. If you are using your AI coding assistant for other work, Otto stays out of the way. Use any of these trigger phrases to activate Otto:

Your team can configure custom trigger phrases during onboarding if you prefer a different convention.

Option 1: VS Code (Recommended)

If you have the Otto VS Code extension installed and your AI coding assistant configured, you do not need to open a specific file or select anything first. Just open the chat panel in your agent and type a trigger phrase:

Hey Otto

Your AI coding assistant sees the Otto MCP tools automatically. Otto will greet you, load your project configuration, and present your current status. No file selection required.

Option 2: CLI

You can also use Otto directly from your agent's CLI in a terminal. The examples below use the Claude Code CLI, but the same approach works with any MCP-compatible agent. This is useful for headless environments, remote servers, or if you prefer working in a terminal.

Single-repo project

Navigate to your project directory and start your agent. Otto's MCP server is already configured in your mcp_servers.json (see Claude Code Setup), so it connects automatically:

cd ~/projects/my-app
claude

Then type a trigger phrase to activate Otto:

Hey Otto

Multi-repo project

Otto licenses per application, and applications often span multiple repos (frontend, API, mobile, infrastructure, etc.). Use the --add-dir flag to give Claude Code access to all of them in a single session:

claude \
  --add-dir ~/projects/my-app-web \
  --add-dir ~/projects/my-app-api \
  --add-dir ~/projects/my-app-mobile

Run this from any of the repos, or from a parent directory. Otto will detect all configured repos and operate across them.

Restricting tool access (optional)

If your organization requires explicit tool allowlisting, you can use the --allowedTools flag to control which Claude Code tools are available during the session:

claude \
  --allowedTools "Bash,Read,Write,Edit,Glob,Grep,WebFetch,WebSearch" \
  --add-dir ~/projects/my-app-web \
  --add-dir ~/projects/my-app-api

This is optional. Without it, Claude Code uses its default tool permissions. Consult the Claude Code documentation for the full list of available tools.

What Happens on Activation

When you say a trigger phrase, Otto runs through its startup sequence:

1. Greet — Otto confirms it is online
2. Load configuration — reads your project settings, repo list, and integrations
3. Health check — verifies git status, stack compliance, and readiness gate across all repos
4. Present status — shows your project name, stack, readiness tier, licensed products, and work queue state
5. Wait for instruction — Otto does not auto-start work; you decide what to do next

If this is the first time Otto runs on your project, it will automatically start the onboarding process to learn about your stack and configure itself.

VS Code vs CLI — Which Should I Use?

Both options connect to the same Otto MCP server, use the same license, and produce the same results. Choose whichever fits your workflow.

Deactivation

Otto automatically steps back when you start working on something unrelated. You can also explicitly deactivate it by saying stop, exit otto, or done with otto. Otto saves your session state and goes silent until you invoke it again.