Overview

This guide walks you through connecting Horizon to an AI provider. Horizon uses AI for triage, smart tagging, and other capabilities. You can bring your own API key from any of the 10 supported providers.

Where to find it: Admin Panel → Integrations → AI → Configuration

Prerequisites

• Admin access to your Horizon account
• An API key from your chosen AI provider (setup instructions for each provider below)

Steps (All Providers)

1. Navigate to Admin Panel → Integrations → AI → Configuration
2. Click Edit to enable the form
3. Select your AI Provider from the dropdown
4. Select or enter your AI Model (some providers show a dropdown, others require you to type the model name)
5. Enter your API Key (not required for Ollama)
6. Fill in any provider-specific fields that appear (see your provider's section below)
7. Click Save

Result

After saving, click the Test Connection button to verify your credentials are valid and Horizon can reach the provider. Once confirmed, Horizon will use your configured provider for all AI capabilities.

Provider Setup Guides

Anthropic

Difficulty: Easy

To get your key:

1. Sign up or log in at console.anthropic.com
2. Add a payment method under Settings → Billing (separate from Claude.ai chat subscriptions)
3. Go to Settings → API Keys, click Create Key
4. Copy immediately — only shown once. Starts with sk-ant-

For more details, see Anthropic's official documentation.

Azure OpenAI

Difficulty: Hard

To set up:

1. You need a pay-as-you-go Azure subscription. Free trial does not include Azure OpenAI.
2. Most models (e.g. gpt-4o) are available immediately. Only restricted models (e.g. GPT-4 vision, abuse-monitoring opt-out) still require a request at aka.ms/oai/access.
3. In Azure Portal, search for "Azure OpenAI" and click Create.
4. Fill in subscription, resource group, region, and name. The name you choose here is your Resource Name.
5. Go to AI Foundry Portal, sign in, and select your project (or create one and attach the resource).
6. Under Deployments, click Create deployment. Select a model (gpt-4o recommended) and give it a name — this is your Deployment ID. Set Tokens per Minute Rate Limit to at least 10K. If you get a "capacity 0" error, request quota at Azure Portal → Quotas → Cognitive Services.
7. Back in Azure Portal, go to your resource → Keys and Endpoint. Copy Key 1 or Key 2 (either works).
8. In Horizon, enter all four values: API Key, Resource Name, Deployment ID, and optionally API Version.

Common gotchas:

• Not all models are available in all Azure regions.
• The Deployment ID is whatever you named the deployment, not the model name.
• Some newer models require a separate access request beyond the initial approval.
• Azure periodically deprecates API versions.
• Free trial subscriptions cannot deploy Azure OpenAI.
• "Capacity 0" deployment errors mean you have no TPM quota — request it at Azure Portal → Quotas → Cognitive Services.

For more details, see Microsoft's official Azure OpenAI documentation.

DeepSeek

Difficulty: Easy

To get your key:

1. Sign up at platform.deepseek.com
2. Deposit funds — prepaid balance model, very cheap
3. Generate key from the API Keys page

For more details, see DeepSeek's official documentation.

Google Gemini

Difficulty: Easy

To get your key:

1. Go to aistudio.google.com/apikey, sign in with Google account
2. Click Create API Key
3. Select or create a Google Cloud project when prompted
4. Copy the key. No billing required for free tier (rate-limited).

For more details, see Google's official Gemini documentation.

Groq

Difficulty: Easy

To get your key:

1. Sign up at console.groq.com (Google, GitHub, or email)
2. Generate key from the Keys page
3. Free tier available, no credit card required. Groq hosts open-source models on their own hardware.

For more details, see Groq's official documentation.

Mistral

Difficulty: Easy

To get your key:

1. Sign up at mistral.ai, verify email
2. Generate key from the API Keys page in the console
3. Free tier available with rate limits. Add payment method for higher usage.

For more details, see Mistral's official documentation.

Ollama (Self-Hosted)

Difficulty: Moderate

To set up:

1. Install Ollama from ollama.com
2. Pull a model (e.g. ollama pull llama3.2)
3. Enter the base URL where Ollama is running

Important: If Horizon is cloud-hosted, Ollama must be network-accessible — localhost won't work. Ollama has no built-in authentication, so put it behind a reverse proxy or VPN if exposing to the internet.

For more details, see Ollama's official documentation.

OpenAI

Difficulty: Easy

To get your key:

1. Sign up or log in at platform.openai.com
2. Add a payment method under Settings → Billing (pay-as-you-go, separate from ChatGPT subscriptions)
3. Go to Settings → API Keys, click Create new secret key
4. Copy immediately — only shown once

For more details, see OpenAI's official documentation.

OpenRouter

Difficulty: Easy

To get your key:

1. Sign up at openrouter.ai (Google, GitHub, or email)
2. Generate key from the Keys page
3. Free tier available for select models. Add credits for paid models.

For more details, see OpenRouter's official documentation.

xAI (Grok)

Difficulty: Easy

To get your key:

1. Sign up at console.x.ai (X/Twitter account or email)
2. Generate key from the console dashboard
3. Free credits may be provided on signup. Add payment method for continued usage.

For more details, see xAI's official documentation.