# FAQ

> Answers to common questions about access, models, data, limits, and billing on Kurrens.

<Faq>

<section data-faq-group>
<h2>Getting started</h2>

<details>
<summary>How do I get an API key?</summary>
<p>Kurrens is in pre-launch and API access is by invitation. <a href="/contact?topic=access">Request access</a> and we'll issue a key. Keys are shown once, when created — store them in an environment variable such as <code>KURRENS_API_KEY</code>.</p>
</details>

<details>
<summary>Is Kurrens compatible with the OpenAI API?</summary>
<p>Yes. Point any OpenAI SDK at <code>https://api.kurrens.ai/v1</code>, use your Kurrens key, and set <code>model</code> to a Kurrens model id. Chat Completions, streaming, tool calling, structured outputs, and vision work as documented. See the <a href="/docs/getting-started/quickstart">Quickstart</a>.</p>
</details>

<details>
<summary>Can I use Kurrens with Cursor, LangChain, LiteLLM, or Dify?</summary>
<p>Yes — anything that lets you set an OpenAI-compatible base URL works. See <a href="/docs/integrations/openai-sdk">Integrations</a>. Tools that only speak the Anthropic Messages API aren't supported.</p>
</details>

<details>
<summary>Is Kurrens on OpenRouter?</summary>
<p>Our catalog is published in the OpenRouter provider format at <a href="/models.json">kurrens.ai/models.json</a>. We'll announce here when Kurrens is live as an OpenRouter provider.</p>
</details>
</section>

<section data-faq-group>
<h2>Models</h2>

<details>
<summary>Which models do you serve?</summary>
<p>Open-weight models from DeepSeek, Moonshot AI (Kimi), Z.ai (GLM), Qwen, MiniMax, Xiaomi, OpenAI (gpt-oss), and Google (Gemma). The full, searchable list is in the <a href="/docs/getting-started/models">Model catalog</a>.</p>
</details>

<details>
<summary>What precision are models served at? Do you quantize them further?</summary>
<p>Each model's precision (for example FP8) is listed in the catalog and in <code>models.json</code>. We never lower the precision behind an existing model id. See <a href="/docs/models/precision-and-versioning">Precision &amp; versioning</a>.</p>
</details>

<details>
<summary>Will the model behind an id change?</summary>
<p>No. A model id always points to the same weights at the same precision. Updated releases get a new id, and old ids are retired with at least 30 days' notice.</p>
</details>

<details>
<summary>Can you add a model?</summary>
<p><a href="/contact?topic=model-request">Request it</a>. We add models based on demand and license terms. For a model only you need, a <a href="/docs/dedicated/overview">dedicated endpoint</a> is usually faster.</p>
</details>

<details>
<summary>Do you support image or video input?</summary>
<p>Yes, on models that support it — for example Kimi K3, MiniMax M3, GLM 5.3 Flash, and Qwen3.8. Filter the <a href="/docs/getting-started/models">Model catalog</a> by "Image / video input" and see <a href="/docs/inference/vision">Vision</a>.</p>
</details>
</section>

<section data-faq-group>
<h2>Data &amp; privacy</h2>

<details>
<summary>Do you store my prompts or completions?</summary>
<p>No. Prompts and completions are processed in memory and discarded when the response completes. We keep request metadata — token counts, latency, status code — for billing. See <a href="/docs/data-security/zero-data-retention">Zero data retention</a>.</p>
</details>

<details>
<summary>Do you train on my data?</summary>
<p>No, never — not our own models, not anyone else's. This is written into our <a href="/legal/data-policy">Data Policy</a>, which takes precedence over our other policies.</p>
</details>

<details>
<summary>Where are my requests processed?</summary>
<p>In Johor (Malaysia) by default, with Jakarta (Indonesia) and Dublin (Ireland) as failover regions. Dedicated endpoints can be pinned to one region. See <a href="/docs/getting-started/regions">Regions &amp; data residency</a>.</p>
</details>

<details>
<summary>Can I sign a DPA?</summary>
<p>Our <a href="/legal/dpa">Data Processing Addendum</a> applies automatically when you accept the Terms. For a countersigned copy, email <a href="mailto:support@kurrens.ai">support@kurrens.ai</a>.</p>
</details>
</section>

<section data-faq-group>
<h2>Limits &amp; reliability</h2>

<details>
<summary>What are the rate limits?</summary>
<p>Limits are per API key, in requests and tokens per minute, and depend on your account and the model. Contact us for higher limits. See <a href="/docs/getting-started/rate-limits">Rate limits</a>.</p>
</details>

<details>
<summary>Why did I get a 429 instead of a slow response?</summary>
<p>We fail fast instead of queueing. When no region has capacity, you get <code>429</code> immediately so your client can retry or fall back without waiting. The OpenAI SDKs retry automatically.</p>
</details>

<details>
<summary>Why does a reasoning model's stream go quiet for a while?</summary>
<p>The model is thinking. We send SSE comment lines (<code>: keep-alive</code>) during long reasoning so connections don't time out — your client should ignore them. See <a href="/docs/inference/streaming">Streaming</a>.</p>
</details>
</section>

<section data-faq-group>
<h2>Billing</h2>

<details>
<summary>How am I billed?</summary>
<p>Per token, at the input, cached-input, and output rates listed for each model on <a href="/pricing">kurrens.ai/pricing</a>. During pre-launch, prices are estimates and may change.</p>
</details>

<details>
<summary>Are reasoning tokens billed?</summary>
<p>Yes, at the output rate. They are reported in <code>usage.completion_tokens_details.reasoning_tokens</code>.</p>
</details>

<details>
<summary>Am I charged if I cancel a streaming request?</summary>
<p>Only for tokens generated before the connection closed.</p>
</details>

<details>
<summary>Do you offer invoicing?</summary>
<p>Yes, for enterprise customers on monthly terms. <a href="/contact?topic=sales">Talk to sales</a>.</p>
</details>
</section>

<section data-faq-group>
<h2>Support</h2>

<details>
<summary>How do I contact support?</summary>
<p>Email <a href="mailto:support@kurrens.ai">support@kurrens.ai</a> and include the request id from the response headers. We reply Monday–Saturday, 09:00–18:00 Singapore time (UTC+8).</p>
</details>

<details>
<summary>How do I report a security issue?</summary>
<p>Email <a href="mailto:support@kurrens.ai?subject=Security%20report">support@kurrens.ai</a> with the subject line "Security report". See <a href="/docs/data-security/security">Security</a>.</p>
</details>
</section>

</Faq>