Skip to main content
The Nozle Agent Kit is a set of skills, guard hooks, and a small checker (nozle-check) for coding agents. With it, an agent can add Nozle billing to your application. The agent checks your stack, reads your Nozle account, studies your code, and asks only the questions that are still open. It then shows a plan and edits only the files you approve. The kit is open source at nozle-dev/nozle-agent-kit.

Quick start

1

Install the kit

For Claude Code:
For Codex CLI, OpenCode, and other hosts, see Install.
2

Open your repository on a clean branch

In Codex, run codex --enable default_mode_request_user_input instead of claude.
3

Ask the agent to integrate Nozle

Say “integrate Nozle”. The agent checks your stack and asks you to log in once.
4

Answer its questions and approve the scope

The agent studies your code and shows the choices that fit, with a recommendation first. It edits only the files you approve.
5

Do your dashboard steps

Create your API keys and put them in your .env. Create anything else your account is missing. Then run nozle-verify. For the full flow, see Run an integration.

Supported hosts

What the agent does and does not do

Does

Inspects your code and your Nozle configuration. Asks focused billing questions, with a recommendation first. Implements the approved integration. Installs the packages it needs with your package manager, after you approve the command. Verifies it with your tests.

Never does

Changes your Nozle account. Creates or reads an API key. Installs a package that the approved scope does not list. Runs your app, or sends a billable event.
The account connection uses OAuth and is read-only on the server. You create plans, meters, and credit systems in the dashboard. The agent tells you which ones you need, and why.

Skills

In Claude Code, the skills are /nozle-agent-kit:<name>. In Codex and OpenCode, name the skill or describe the task.

Next steps

Install and log in

Install the kit for your host, log in, update, or remove it.

Run an integration

Where to start the agent, what it asks, and what to do after it finishes.