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

# AI Provider

> Choose which LLM powers Ankra's AI features - the managed default, your own OpenRouter or Anthropic key, or any OpenAI-compatible endpoint.

Organisation admins choose which model provider powers all AI features (AI Assistant, AI Insights, incident analysis, CI/CD file generation, summaries). Configure it under **AI** → **Settings** → **Models**.

## Provider Modes

| Mode                  | What it is                                                                | When to use it                                                                                                                 |
| --------------------- | ------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------ |
| **Ankra** (default)   | Platform-managed key, zero configuration                                  | Getting started; no key management. Usage counts against the [free monthly AI allowance](/platform/ai-usage)                   |
| **Anthropic**         | Your own Anthropic API key                                                | Your own billing and data agreement, with every tier served by Claude                                                          |
| **OpenRouter**        | Your own OpenRouter API key                                               | Unlimited AI on your own billing with the same default models and failover chain - the recommended way past the free allowance |
| **OpenAI-compatible** | Any OpenAI v1 endpoint: LiteLLM, vLLM, TGI, Ollama, Together, and similar | Self-hosted or third-party models behind an OpenAI-style API                                                                   |

Credentials and the active selection are stored separately - you can save and validate a configuration first, then switch the active provider when ready.

## Default Models

On the Ankra provider, every organisation is served the same three tiers. The chat model picker selects between them, and **Auto** routes each question to the right one - short questions to Quick, everyday work to the default, and architecture, security, or image questions to Expert.

| Tier                 | Model            | Context     | Notes                                             |
| -------------------- | ---------------- | ----------- | ------------------------------------------------- |
| **Expert** (default) | Kimi K3          | 1M tokens   | Reasoning and image input. What Auto escalates to |
| **Think**            | GLM 5.2          | 1M tokens   | Reasoning at high effort, text only               |
| **Quick**            | Claude Haiku 4.5 | 200k tokens | Fastest and cheapest, for simple questions        |

Because GLM 5.2 does not accept images, Auto sends any question with an attachment to Expert.

These are defaults, not limits. Under **AI** → **Settings** → **Models** you can edit any tier, point it at a different model, add your own, or reset the catalog back to the Ankra defaults. Switching the provider to Anthropic serves every tier with Claude instead.

## Configure Anthropic

<Steps>
  <Step title="Save the key">
    In the **Anthropic** card, paste an API key (must start with `sk-ant-`). Ankra validates it with a one-token health check before saving.
  </Step>

  <Step title="Switch the provider">
    Select **Anthropic** in the provider picker. Switching is blocked until a valid key is saved.
  </Step>
</Steps>

Keys are stored in Vault; the UI only ever shows a masked preview (`sk-ant-...XXXX`). Deleting the key while Anthropic is active automatically reverts the organisation to the Ankra provider.

## Configure OpenRouter

With your organisation's own OpenRouter API key active, Ankra serves the same default models and automatic failover chain through your OpenRouter account - including background and autonomous AI sessions - and your usage never draws on the [free monthly AI allowance](/platform/ai-usage). The key is validated live against OpenRouter at save time and stored only in Vault. Follow the step-by-step guide: [Bring Your Own OpenRouter API Key](/guides/openrouter-api-key).

## Configure an OpenAI-Compatible Endpoint

<Steps>
  <Step title="Save the configuration">
    In the **OpenAI-compatible** card, provide:

    * **Base URL** - the endpoint's OpenAI `/v1` root, `http(s)://` (e.g. `https://my-litellm.internal/v1`)
    * **Model** - the model identifier your endpoint serves
    * **API key** - the endpoint's key

    Ankra probes `GET {'{base_url}'}/models` on save to verify the endpoint responds. An incorrect model id only surfaces at the first AI call.
  </Step>

  <Step title="Switch the provider">
    Select **OpenAI-compatible** in the provider picker.
  </Step>
</Steps>

<Note>
  In OpenAI-compatible mode, tool-using AI paths - the agentic AI Assistant, AI Insights, CI/CD file generation, and deploy analysis - fall back to the Ankra platform provider. Stack descriptions and summaries run natively against your endpoint. The **What runs where** matrix on the settings page shows the current mapping.
</Note>

## Permissions

Only organisation **admins** can view previews, save credentials, or switch providers. Other members see the settings read-only.

## Related Settings

The **AI** settings page has three more tabs:

* **Skills** - organisation-specific AI skills; see [AI Skills](/platform/ai-skills)
* **Insights** - the org-wide toggle for [Proactive AI Insights](/platform/ai-insights)
* **Advanced** - the read-only system-prompt variant in use