# Get started

Point your app at ProxyLLM and make your first request in about five minutes.

*ProxyLLM docs · https://proxyllm.ai/docs/get-started*

ProxyLLM is an OpenAI-compatible gateway. Your app talks to `https://api.proxyllm.ai/v1` exactly the way it talks to OpenAI, and ProxyLLM routes each request to the backend you pick: your Codex subscription, your own Claude Code bridge, or provider API keys.

## 1. Create an account

Sign up at [proxyllm.ai/signup](https://proxyllm.ai/signup). The dashboard is where you connect subscriptions and create keys.

## 2. Connect something to route to

Pick at least one backend:

- **Codex** - bring your ChatGPT/Codex subscription. See [Connect Codex](/docs/codex).
- **Claude Code** - run inference on your Claude subscription through a server you own. See [Claude Code Bridge](/docs/claude-code).
- **Provider API keys** - save OpenAI, Anthropic, Gemini, or other provider keys under [Providers](https://proxyllm.ai/dashboard/providers).

## 3. Create a routing key

A routing key is what your app uses as its API key. Create one under [Routing keys](https://proxyllm.ai/dashboard/routing-keys), add one or more lanes to it, and copy the key. It starts with `pllm_` and is shown once. [How routing works](/docs/routing-keys).

## 4. Point your app at the gateway

Set two environment variables:

```bash
export OPENAI_BASE_URL="https://api.proxyllm.ai/v1"
export OPENAI_API_KEY="pllm_your_routing_key"
```

Then make a request:

```bash
curl https://api.proxyllm.ai/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "flagship",
    "messages": [{"role": "user", "content": "Say hello."}]
  }'
```

`flagship` is a stable alias that always resolves to the best model your key can reach, so it keeps working when providers rotate model names. Every OpenAI SDK and tool that accepts a base URL works; see [Integrations](/integrations) for copy-paste setups.

## 5. Watch it run

Every request shows up under [Usage](https://proxyllm.ai/dashboard/usage) with tokens, cost, and which lane served it. Agents and scripts can read the same stats through the [API](/docs/api) with a read-only account token.