Integrations
Integrations Overview
The gateway sits between your applications and your model providers, so an integration connects one of its two sides: a provider that the gateway calls, or a client that calls the gateway. Each guide shows the configuration and one request that proves the connection works.
Before you start
- A working gateway from the Quickstart.
- Provider access for that gateway. See Connect Your Model Provider.
Select an integration
Model providersSend requests to OpenAI, Azure OpenAI, Vertex AI, Mistral, Groq, and other providers.Application SDKsSend requests from Node.js, Python, Java, .NET, and agent frameworks.IDE clientsRoute editor model requests through the gateway, or connect an agent to MCP tools.Open WebUIConnect a self-hosted chat interface with a client API token.SIEM streamingStream AI usage records to Splunk, Elasticsearch, Datadog, or an HTTPS collector.Gateway APIThe gateway routes for text, embeddings, audio, and moderation.
Model providers
Each provider page shows a target for policy-config.yaml, a test request, and the request families that the adapter supports.
| Group | Providers |
|---|---|
| Frontier labs | OpenAI, Anthropic, Cohere, DeepSeek, Mistral, xAI, Perplexity, Alibaba Qwen |
| Clouds | Amazon Bedrock, Azure OpenAI, Google Vertex AI, Google AI Studio, Amazon SageMaker, Databricks, Cloudera, Cloudflare Workers AI, IBM watsonx.ai |
| Fast inference | Groq, Cerebras, SambaNova, Fireworks, Together AI |
| Routers and gateways | OpenRouter, Vercel AI Gateway, Cloudflare AI Gateway, AIML API, GitHub Models |
| Other | Hugging Face, Replicate, Voyage AI, QuiverAI |
Provider Configuration describes the target fields that all of them share.
Next steps
- Policy Configuration Overview: add rules to the requests the gateway receives.
- Debug AI Requests: trace a request by its ID.
- Troubleshooting: find why a request failed.