Compatible AI providers

OpenClaw can connect to multiple AI model providers. Pick one you already use, or start with our recommended default.

Some providers require accounts or API keys, and availability or pricing can change over time.

Our recommended provider for performance/cost optimization

Recommended: z.ai GLM-4.7

Frontier-level agentic coding performance with subscription economics that fit OpenClaw's heavy workflows.

Read why we recommend GLM-4.7 (we're currently evaluating GLM-5.0 and will update this in a few days)

Most common providers

OpenAI

Broad model ecosystem, common default choice.

Anthropic

Popular for safety-oriented assistant workflows.

Google Gemini

Google model family for chat and coding tasks.

Amazon Bedrock

AWS-managed model access across providers.

OpenRouter

Unified API for many hosted model providers.

z.ai (GLM)

High-performance agentic coding with long context.

Groq

Fast inference focused on low-latency responses.

Mistral

Open and proprietary model options.

Ollama (local)

Run compatible models on your local machine.

Other supported providers

OpenAI Codex

OpenAI coding-focused tooling and models.

GitHub Copilot

AI coding assistant integrated into developer tools.

Google Vertex AI

Google Cloud platform for model deployment.

xAI (Grok)

Grok model access from xAI.

Cerebras

High-throughput inference platform for large models.

Moonshot AI (Kimi)

Kimi models with long-context capabilities.

MiniMax

General-purpose model platform for agents and chat.

Baidu Qianfan

Baidu enterprise platform for model access.

Vercel AI Gateway

Routing layer for multiple model APIs.

LM Studio (OpenAI-compatible endpoint)

Local model serving with OpenAI-compatible endpoints.

vLLM (OpenAI-compatible endpoint)

High-throughput serving engine for compatible APIs.

LiteLLM (OpenAI-compatible proxy)

Unified proxy to standardize multi-provider access.

Last updated: February 2026