# Quickstart

> An empty folder, one command, and four prompts to your coding agent, to an agent that answers.

From an empty folder to an agent that answers. You run one command, and your coding agent does the rest.

**You need** Node 22.12 or later, Docker running, and a coding agent such as Claude Code. Kindgi's runtime image needs access too: sign in at [access.kindgi.com](https://access.kindgi.com/) with GitHub, and run the `kindgi auth registry` command it shows. [Install](https://docs.kindgi.com/start/install.md) has the details.

## 1. Make the project

In an empty folder:

```sh
npx --yes @kindgi/cli init my-agents
```

It makes the project in a new folder, `my-agents`, and gives your coding agent Kindgi's skills, in `my-agents/.claude/skills/` and `my-agents/.agents/skills/`:

```text excerpt
…
✓ Pack scaffolded at …/my-agents/
✓ package.json pins pnpm@10.28.0 (packageManager)
…
```

## 2. Ask your coding agent

Open `my-agents` in your coding agent. Send it these prompts, one at a time:

1. `Install this project's dependencies and start Kindgi in the background.`
2. `Connect a real model.`
3. `Write an agent that reads a customer message and says how urgent it is, and why.`
4. `Run it with: "My order hasn't arrived and I need it tomorrow."`

At the second prompt, your coding agent asks which AI model provider you use. Then it gives you a command that stores the provider's API key, to run yourself, in a terminal in `my-agents`. For Anthropic, it's:

```sh
pnpm exec kindgi secrets set ANTHROPIC_API_KEY --env=local --scope=tenant
```

For another provider, it's the same command with that provider's key name, such as `OPENAI_API_KEY`. It asks for the key twice, and shows it neither time, so the key never goes into the chat:

```text
Value:
Re-enter to confirm:
```

If your coding agent asks you to paste the key into the chat, or to put it in a file, run the command instead. Then tell your coding agent it's done: it registers the provider, with `kindgi providers register` and the provider's preset, such as `--preset=anthropic`.

## Find the console

The console shows each run: what the agent was asked, the steps it took and what it answered. When Kindgi starts, it prints the console's address:

```text
    Console    http://127.0.0.1:4000/console/   (open in your browser)
```

Your coding agent started it, so ask it: `What's the console's address?` The port is `4000`, or the next free one. On the sign-in page, click **Sign in as seeded user**.

## Next

- [Your coding agent](https://docs.kindgi.com/start/coding-agents.md): what else to ask it, and how it knows Kindgi.
- [Quickstart: TypeScript](https://docs.kindgi.com/start/quickstart-typescript.md): the same steps, by hand. Also in [Python](https://docs.kindgi.com/start/quickstart-python.md), [Java](https://docs.kindgi.com/start/quickstart-java.md) and [Scala](https://docs.kindgi.com/start/quickstart-scala.md).
- [Add Kindgi to an existing app](https://docs.kindgi.com/start/existing-app.md).