Getting Started with the Optibot Claude Code Skill

The Optibot Claude Code skill brings AI-powered code reviews directly into your Claude Code sessions. Instead of switching between tools, you can trigger a full Optibot review from within Claude and see results inline as you work.

Overview

The skill extends Claude Code with Optibot’s code review capabilities. Once installed, you can ask Claude to review your local changes, a branch diff, or a specific diff file — and Optibot returns a review with findings, suggestions, and security checks. Ask Claude to review changes it just made itself and it uses agent mode, which returns structured findings it acts on directly.

Requirements

  • Claude Code installed and running
  • An Optibot account at agents.getoptimal.ai
  • Node.js 20 or later, because the skill runs the Optibot CLI
  • The Optibot CLI. You do not have to install it yourself: the skill checks for it and runs npm install -g @optimalai/optibot when it is missing. That install succeeds on an older Node, but the CLI then fails at runtime, so check node --version first if a freshly installed optibot errors on its first command

Installation

Option 1: Inside a Claude Code Session

  1. Open Claude Code and start a new session.
  2. Type /plugin to open the Plugin Marketplace.
  3. Browse the Marketplace tab.
  4. Add Optimal-AI/optibot-skill to find the Optibot plugin.
  5. Click Discover to load the plugin listing.
  6. Click Install to install the skill.
  7. Run /reload-plugins to activate the newly installed skill.

Option 2: CLI Terminal

Run the following commands in your terminal:

claude plugin marketplace add Optimal-AI/optibot-skill

Then install it:

claude plugin install optibot@optimal-ai

Verifying Installation

After installing, type /plugin inside Claude Code and confirm that optibot appears in your installed plugins list.

Logging In

Authenticate with your Optibot account:

optibot login

This opens your browser and prompts you to log in via your Optibot account. It stores a 90-day session token in ~/.optibot/config.json, and the skill uses it automatically from then on.

That session token is not an API key. An API key is the separate, long-lived optk_ credential you mint with optibot setup ci for CI pipelines, and it stays bound to the organization that was active when you created it.

How It Works

The skill is a set of instructions, not a set of tools. It tells Claude when to reach for Optibot and which command to run, and it grants Claude permission to run the optibot CLI. When you ask Claude to review code, Claude runs a command such as optibot review in your shell, and the CLI submits the diff to Optibot’s review engine and prints the result back into the session.

If you want Optibot as native tools in an editor instead, that is the Optibot MCP server, a separate package with its own setup.

Reviews cover:

  • Security vulnerabilities
  • Logic errors
  • Code quality and maintainability
  • Test coverage gaps
  • Performance concerns

How to Use Optibot with Claude

Once the skill is installed, use natural language inside Claude Code:

What you sayWhat happens
”review my changes”Reviews your uncommitted local changes
”review my branch”Reviews all commits on your current branch vs. main
”review this diff”Reviews a provided diff or patch file
”set up optibot”Walks you through Optibot configuration for a new repo
”create an API key for CI”Generates an Optibot API key for use in CI/CD pipelines
”review the changes you just made”Runs an agent review: a fast pass returning structured findings Claude acts on directly

💡 You can combine these with Claude’s other capabilities: “Review my changes and explain any security findings in detail.”

Updating

Refresh the marketplace listing, then update the plugin:

claude plugin marketplace update optimal-ai
claude plugin update optibot@optimal-ai

Restart Claude Code afterwards. The new version is on disk right away, but the session you ran the command in keeps the version it loaded at startup.

The skill runs the Optibot CLI, which is a separate package on its own version number, so update it in the same pass:

npm install -g @optimalai/optibot@latest

To see which versions you are on, run claude plugin list for the skill and optibot --version for the CLI.

If you installed the skill before April 2026, it may have set up the older optibot package. See Upgrading from the optibot package.

Keeping the two in step matters because new skill behaviour lands behind a CLI minimum. Agent review mode needs CLI 0.8.0 or later; on an older CLI the --agent flag comes back as unknown option.

Uninstalling

To remove the Optibot skill from Claude Code:

claude plugin uninstall optibot

Then run /reload-plugins inside Claude Code to confirm it has been removed.