Providers Overview

AI provider system architecture and 50+ supported services

NullClaw supports 50+ AI providers through a unified vtable-based interface. This enables seamless model switching without changing your agent configuration.

Provider Interface

All providers implement the Provider vtable interface defined in src/providers/root.zig:

pub const Provider = struct {
    ptr: *anyopaque,
    vtable: *const VTable,

    pub const VTable = struct {
        chatWithSystem: *const fn (...) anyerror![]const u8,
        chat: *const fn (...) anyerror!ChatResponse,
        supportsNativeTools: *const fn (ptr: *anyopaque) bool,
        getName: *const fn (ptr: *anyopaque) []const u8,
        deinit: *const fn (ptr: *anyopaque) void,
        warmup: ?*const fn (ptr: *anyopaque) void = null,
        supports_streaming: ?*const fn (ptr: *anyopaque) bool = null,
        supports_vision: ?*const fn (ptr: *anyopaque) bool = null,
        stream_chat: ?*const fn (...) anyerror!StreamChatResult = null,
    };
};

Core Providers

NullClaw includes dedicated implementations for these major providers:

• OpenAI — GPT-4o, GPT-5, o1, o3, o4
• Anthropic — Claude 4, Sonnet, Opus
• OpenRouter — 200+ model aggregator
• Ollama — Local LLMs (Llama, Mistral, Qwen)
• Gemini — Google Gemini 2.0, 1.5 Pro
• Claude CLI — Reuses ~/.claude/ credentials
• Codex CLI — GitHub Copilot integration
• OpenAI Codex — Legacy Codex models

Compatible Providers (41)

These providers use the OpenAI-compatible API format:

Major Cloud Providers

• Groq, Mistral, DeepSeek, xAI, Cerebras, Perplexity, Cohere

Gateways & Aggregators

• Venice, Vercel AI Gateway, Together AI, Fireworks AI, Hugging Face
• AIHubMix, SiliconFlow, Chutes, Synthetic, Poe

China Providers

• Moonshot (Kimi), GLM (Zhipu), Z.AI, MiniMax, Qwen, Qianfan, Doubao

Infrastructure

• Amazon Bedrock, Cloudflare AI Gateway, GitHub Copilot, NVIDIA NIM, OVHcloud

Local Servers

• LM Studio, vLLM, llama.cpp, SGLang, Osaurus, LiteLLM

Provider Selection

Select a provider in ~/.nullclaw/config.json:

{
  "provider": "anthropic",
  "model": "claude-sonnet-4",
  "api_key": "sk-ant-..."
}

For OpenRouter:

{
  "provider": "openrouter",
  "model": "openrouter/anthropic/claude-sonnet-4",
  "api_key": "sk-or-..."
}

For local Ollama:

{
  "provider": "ollama",
  "model": "llama3.2",
  "base_url": "http://localhost:11434"
}

Custom Endpoints

Use custom: prefix for arbitrary OpenAI-compatible endpoints:

{
  "provider": "custom:https://api.example.com/v1",
  "model": "my-model",
  "api_key": "..."
}

Use anthropic-custom: for Anthropic-format endpoints:

{
  "provider": "anthropic-custom:https://proxy.example.com",
  "model": "claude-sonnet-4"
}

Capabilities

Capability	Description
supportsNativeTools()	Function calling / tool use
supportsStreaming()	SSE streaming responses
supportsVision()	Image/multimodal input
warmup()	Pre-warm TLS connections

Provider-Specific Notes

• OpenAI: Supports reasoning models (o1, o3, gpt-5) with reasoning_effort parameter
• Anthropic: OAuth tokens (sk-ant-oat01-) use Bearer auth instead of x-api-key
• Gemini: Supports API keys, OAuth, and Gemini CLI (~/.gemini/oauth_creds.json)
• Ollama: No authentication required for local servers
• OpenRouter: Requires HTTP-Referer and X-Title headers

Next Steps

• OpenAI Configuration
• Anthropic Configuration
• OpenRouter Configuration
• Ollama Local Setup
• Gemini Configuration