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

# Boole AI Cloud API: OpenAI-Compatible Managed Inference

> Connect to the Boole AI cloud API — OpenAI-compatible endpoints, zero-retention policy, SOC 2 Type II certified, and $20 free credits to start.

The Boole AI cloud API lets you run inference without any GPU hardware. It uses the same OpenAI-compatible interface as the local binary — just swap the base URL and add your API key. Every model available locally is also available via the cloud API, at the same slugs, with the same request and response shapes.

## Endpoint

All API requests go to:

```
https://api.boole.dev/v1
```

Every public endpoint is available under this base URL — `/chat/completions`, `/completions`, `/audio/transcriptions`, and `/models`. Point any OpenAI-compatible SDK at this URL and it works without further changes.

## Authentication

Include your API key in the `Authorization` header on every request:

```bash theme={null}
Authorization: Bearer $BOOLE_API_KEY
```

Store your key in an environment variable rather than hardcoding it. Retrieve or rotate your keys at any time from [your API keys page](/account/api-keys).

<Warning>
  Never commit your API key to source control. Treat it like a password — if it is exposed, rotate it immediately from the API keys page.
</Warning>

## Quick Test

Run this curl command to confirm your key works and the endpoint is reachable:

```bash theme={null}
curl https://api.boole.dev/v1/chat/completions \
  -H "Authorization: Bearer $BOOLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama-3.3-70b-instruct",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
```

A `200` response with a `choices` array confirms everything is set up correctly.

## Cloud vs. Local

Choose the deployment mode that fits your latency, privacy, and infrastructure requirements. The two modes are fully interchangeable at the API level.

|                       | Cloud API               | Local Binary         |
| --------------------- | ----------------------- | -------------------- |
| **Setup time**        | Instant                 | \~4 minutes          |
| **Hardware required** | None                    | NVIDIA GPU           |
| **Cost**              | Per token (see Pricing) | Free (GPU amortized) |
| **Data privacy**      | Zero-retention policy   | Fully local          |
| **Offline capable**   | No                      | Yes                  |
| **Cold start (TTFT)** | \~62 ms                 | \~380 ms             |

<Accordion title="What does zero-retention mean for the cloud API?">
  Boole AI does not log, store, or train on your prompts or completions. Requests are processed in memory and discarded immediately after the response is sent. Boole AI is SOC 2 Type II certified — the zero-retention policy is an audited control, not just a stated intent.
</Accordion>

## Pricing

The cloud API is billed per token. A few example prices:

| Model                  | Input             | Output            |
| ---------------------- | ----------------- | ----------------- |
| Llama 3.3 70B Instruct | \$0.09 / M tokens | \$0.14 / M tokens |
| Qwen 2.5 72B           | \$0.11 / M tokens | \$0.16 / M tokens |
| Mixtral 8×22B          | \$0.13 / M tokens | \$0.19 / M tokens |
| DeepSeek V3            | \$0.08 / M tokens | \$0.12 / M tokens |
| Whisper Large v3       | \$0.02 / min      | —                 |

New accounts receive **\$20 in free credits** (approximately 1 million tokens) with no credit card required. See the [full pricing page](/concepts/pricing) for all 42 models.

<Tip>
  The cloud API and local binary are fully interchangeable. Switch between them by changing a single URL — no other code changes needed. Use the cloud API during development or when you need to burst beyond local GPU capacity, and the local binary when you need offline access or full data isolation.
</Tip>