Providers

Every provider has its own SDK, its own auth story, and its own feature set. The table below is a quick cross-reference; the provider pages have the details and copy-pasteable code.

Supported providers

Provider	Client adapter	Recommended run adapter	Notes
OpenAI	`openaiClientAdapter`	`openaiResponsesRunAdapter`	Native Responses API
Anthropic	`anthropicClientAdapter`	`completionsRunAdapter`	Claude via `/chat/completions`
Google (Gemini)	`googleClientAdapter`	`geminiRunAdapter`	Native Google tools
Azure OpenAI	`azureOpenaiClientAdapter`	`azureResponsesRunAdapter`	Responses on Azure
Azure AI Foundry	`azureAiProjectClientAdapter`	`azureAgentsRunAdapter`	Stateful Agents on Azure
Mistral	`mistralClientAdapter`	`completionsRunAdapter`	Mistral Large/Small
Groq	`groqClientAdapter`	`completionsRunAdapter`	Ultra-fast inference
Together	`togetherClientAdapter`	`completionsRunAdapter`	Llama, Qwen, GPT-OSS
OpenRouter	`openRouterClientAdapter`	`completionsRunAdapter`	300+ models
Perplexity	`perplexityClientAdapter`	`completionsRunAdapter`	Sonar, web-grounded
Ollama	`ollamaClientAdapter`	`completionsRunAdapter`	Local models

Provider guides

OpenAIResponses and Assistants APIs, native tools, and OpenAI-managed state.

AnthropicClaude models through /chat/completions.

GoogleGemini with native Google tooling.

AzureAzure OpenAI and Azure AI Foundry Agents.

MistralMistral Large, Small, and Codestral.

GroqUltra-fast inference for Llama, Qwen, and GPT-OSS.

TogetherOpen-weight models hosted on Together AI.

OpenRouter300+ models behind a single OpenAI-compatible endpoint.

PerplexitySonar — live web-grounded answers with citations.

OllamaRun local open-weight models via an Ollama instance.