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

# LLM Support

> Supported AI providers for LLM-driven policy evaluation

<Note>
  Bring your own LLM provider key is only available on Chainloop's Enterprise [paid plan](https://chainloop.dev/pricing).
</Note>

## Overview

Chainloop integrates with Large Language Model (LLM) providers to power AI-driven policy evaluation, data analysis and more.

AI providers can be registered from the [Integrations page](https://app.chainloop.dev/integrations?filterType=ai-provider) in your organization. Once configured, they are automatically used by [LLM-driven policies](/guides/llm-policies) and the [`chainloop.evidence_prompt`](/reference/builtin-functions#chainloopevidence_prompt) builtin function.

<img src="https://mintcdn.com/chainloop/i7LPwzemSBO4bbcz/reference/img/llm-support-ai-providers.png?fit=max&auto=format&n=i7LPwzemSBO4bbcz&q=85&s=dfa961fca5e8cb2ad34b5843ac35a7c0" alt="AI Provider integrations" width="2146" height="462" data-path="reference/img/llm-support-ai-providers.png" />

## Supported Providers

| Provider      | Model             | Type                             | Description                                          |
| ------------- | ----------------- | -------------------------------- | ---------------------------------------------------- |
| **Anthropic** | Claude Sonnet 4.5 | Direct API                       | Connect directly to Anthropic's Claude API           |
| **OpenAI**    | GPT-4.1           | Direct API                       | Connect directly to OpenAI's API                     |
| **OpenAI**    | GPT-4.1           | Azure OpenAI (Microsoft Foundry) | Connect to OpenAI models hosted on Microsoft Foundry |

<Warning>
  If you have both Anthropic and OpenAI registered, Anthropic will take precedence and be used as the default provider for LLM evaluations.
</Warning>

## Anthropic

Register an Anthropic integration by providing your API key. Chainloop will use it to send evidence and prompts to Anthropic's Claude models for evaluation.

**Registration fields:**

| Field       | Required | Description                         |
| ----------- | -------- | ----------------------------------- |
| Name        | Yes      | A unique name for this registration |
| Description | No       | Optional description                |
| Api Key     | Yes      | Your Anthropic API key              |

## OpenAI

Register an OpenAI integration by providing your API key. This connects directly to OpenAI's API.

**Registration fields:**

| Field       | Required | Description                         |
| ----------- | -------- | ----------------------------------- |
| Name        | Yes      | A unique name for this registration |
| Description | No       | Optional description                |
| Api Key     | Yes      | Your OpenAI API key                 |

## OpenAI on Microsoft Foundry

The OpenAI integration also supports connecting to Azure-hosted OpenAI models through Microsoft Foundry. To use Azure OpenAI, provide a custom endpoint pointing to your Azure OpenAI deployment alongside your API key.

**Registration fields:**

| Field       | Required | Description                                                                   |
| ----------- | -------- | ----------------------------------------------------------------------------- |
| Name        | Yes      | A unique name for this registration (e.g., "Azure OpenAI")                    |
| Description | No       | Optional description                                                          |
| Endpoint    | No       | Your Azure OpenAI endpoint URL (e.g., `https://<resource>.openai.azure.com/`) |
| Api Key     | Yes      | Your Azure OpenAI API key                                                     |

<Tip>
  The **Endpoint** field is what distinguishes a direct OpenAI registration from an Azure OpenAI one. When an endpoint is provided, Chainloop routes requests to your Azure deployment instead of the OpenAI API directly.
</Tip>

## Related Resources

* [How to run LLM-driven policies](/guides/llm-policies) — Guide on using AI-powered prompts to evaluate evidence
* [Builtin functions reference](/reference/builtin-functions#chainloopevidence_prompt) — The `chainloop.evidence_prompt` function for custom Rego policies
* [Integrations](/concepts/integrations) — General integration setup instructions