> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tulipai.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart

> Follow the Tulip user flow from wallet connection to paid model inference.

Tulip uses Robinhood Chain for credits, markets, permits, and settlement. Inference requests travel through the Tulip Gateway to an eligible provider endpoint.

<Info>
  This guide covers Tulip v1's production flow on Robinhood Chain mainnet. Version 1 routes inference through verified Hugging Face Inference Endpoints; later versions may support additional vendors.
</Info>

## Prerequisites

You need:

* An EVM wallet that supports Robinhood Chain.
* Native gas on Robinhood Chain.
* USDG if you plan to buy model credits.
* A Tulip Gateway API key if you plan to call the Gateway directly.

## Complete your first inference request

<Steps>
  <Step title="Connect to Robinhood Chain">
    Add Robinhood Chain to your wallet and confirm chain ID `4663`. See [Network and addresses](/reference/addresses).
  </Step>

  <Step title="Choose a model market">
    Open **Explore**, search by model name or Hugging Face repository, and verify the pinned revision before trading.
  </Step>

  <Step title="Buy model credits">
    Open the model's market, enter the USDG amount, review the quote and minimum received, then approve and submit the swap.

    The swap executes through Permit2 and the Universal Router against the canonical Uniswap v4 pool.
  </Step>

  <Step title="Create an inference permit">
    Open **Credits**, select a credit held by your wallet, choose how much to make available, and approve the escrow deposit.

    A permit is bound to one model. You may leave provider selection open or pin the permit to one provider offer.
  </Step>

  <Step title="Send a request">
    Open **Chat**, select the model, choose the permit, and send a message. The gateway verifies the permit, selects an eligible provider, reserves the maximum quoted charge, and forwards the request.
  </Step>

  <Step title="Review the charge">
    The provider response includes Tulip request, quote, offer, and tariff identifiers. After metering, the gateway signs a usage receipt for on-chain settlement.
  </Step>
</Steps>

<Check>
  A completed request is charged in the model's credit, not in USDG. Any unused reserved amount remains available in the permit.
</Check>

## Use the Gateway directly

The Gateway accepts an OpenAI-compatible chat request with Tulip authorization headers.

```bash theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
curl "$TULIP_GATEWAY_URL/v1/chat/completions" \
  -H "Authorization: Bearer $TULIP_API_KEY" \
  -H "X-Tulip-Permit: $PERMIT_ID" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "owner/repository",
    "messages": [{"role": "user", "content": "Explain this model."}],
    "max_completion_tokens": 256
  }'
```

Set `TULIP_GATEWAY_URL` to the production base URL provided with your Tulip API credentials.

## Next steps

<Columns cols={2}>
  <Card title="Trade model credits" icon="arrow-left-right" href="/guides/trade-credits">
    Understand approvals, quotes, price impact, and transaction flow.
  </Card>

  <Card title="Inference permits" icon="badge-dollar-sign" href="/concepts/inference-permits">
    Learn how deposits, reservations, settlement, and refunds work.
  </Card>
</Columns>