Skip to main content
OneClaw supports 6 LLM providers with automatic fallback chains. All providers use a unified [provider] configuration section.

​
Supported Providers

ProviderIDDefault ModelAPI Key Env Var
Anthropic Claudeanthropicclaude-sonnet-4-20250514ANTHROPIC_API_KEY
OpenAI GPTopenaigpt-4oOPENAI_API_KEY
DeepSeekdeepseekdeepseek-chatDEEPSEEK_API_KEY
Groqgroqllama-3.3-70b-versatileGROQ_API_KEY
Google Geminigoogle or geminigemini-2.0-flashGOOGLE_API_KEY
Ollama (local)ollamallama3.2:1b(none)

​
Basic Configuration

Minimal provider configuration: 
[provider]
primary = "anthropic"
model = "claude-sonnet-4-20250514"
API key via environment variable: 
export ANTHROPIC_API_KEY="sk-ant-..."

​
Provider Selection

Set the primary field to choose your LLM provider: 
[provider]
primary = "openai"          # OpenAI GPT-4o
model = "gpt-4o-mini"       # Override default model
max_tokens = 2048
temperature = 0.5

[provider]
primary = "deepseek"        # DeepSeek (cost-effective)
model = "deepseek-chat"
max_tokens = 4096
temperature = 0.3

[provider]
primary = "groq"            # Groq (fast inference)
model = "llama-3.3-70b-versatile"
max_tokens = 8192
temperature = 0.7

[provider]
primary = "google"          # Google Gemini
model = "gemini-2.0-flash"
max_tokens = 1024
temperature = 0.4

[provider]
primary = "ollama"          # Local Ollama (no API key needed)
model = "llama3.2:3b"
max_tokens = 2048
temperature = 0.6

​
Model Configuration

​
Available Models by Provider

Anthropic Claude:
  • claude-sonnet-4-20250514 — Best balance (default)
  • claude-haiku-4-5-20251001 — Fast, cheap, good for classification
  • claude-opus-4-5-20250918 — Max quality, expensive
OpenAI GPT:
  • gpt-4o — Latest GPT-4 (default)
  • gpt-4o-mini — Faster, cheaper
DeepSeek:
  • deepseek-chat — General purpose (default)
  • deepseek-reasoner — Advanced reasoning
Groq:
  • llama-3.3-70b-versatile — Llama 3.3 70B (default)
  • mixtral-8x7b-32768 — Mixtral MoE
Google Gemini:
  • gemini-2.0-flash — Fast, efficient (default)
Ollama:
  • llama3.2:1b — Smallest, fastest (default)
  • llama3.2:3b — Better quality
  • qwen2.5:3b — Good multilingual support
  • (Any Ollama-supported model)

​
Model Parameters

[provider]
primary = "anthropic"
model = "claude-sonnet-4-20250514"
max_tokens = 1024          # Maximum response length
temperature = 0.3          # 0.0 (deterministic) to 1.0 (creative)
  • max_tokens: Controls response length (default: 1024)
  • temperature: Controls randomness (default: 0.3)
    • 0.0 — Deterministic, consistent
    • 0.3-0.5 — Balanced (recommended for agents)
    • 0.7-1.0 — Creative, varied

​
Fallback Chains

OneClaw supports automatic failover to backup providers: 
[provider]
primary = "anthropic"
model = "claude-sonnet-4-20250514"
fallback = ["openai", "ollama"]  # Try in order if primary fails
max_retries = 2                   # Retry each provider N times

​
How Fallback Works

  1. Try primary provider (with max_retries attempts)
  2. If primary fails, try fallback[0] (with max_retries)
  3. If fallback[0] fails, try fallback[1], etc.
  4. If all providers fail, return error

​
Example: Cloud with Local Fallback

[provider]
primary = "anthropic"              # Cloud primary
model = "claude-sonnet-4-20250514"
fallback = ["ollama"]              # Local fallback if network fails
max_retries = 1

[provider.keys]
anthropic = "sk-ant-..."
Ollama runs locally on http://localhost:11434 (no API key needed).

​
Example: Multi-Cloud Fallback

[provider]
primary = "anthropic"              # Primary: Claude
model = "claude-sonnet-4-20250514"
fallback = ["openai", "deepseek"] # Fallback chain
max_retries = 2                    # 2 retries per provider

[provider.keys]
anthropic = "sk-ant-..."
openai = "sk-..."
deepseek = "sk-..."

​
API Key Configuration

Use the [provider.keys] table for multiple providers: 
[provider]
primary = "anthropic"
model = "claude-sonnet-4-20250514"
fallback = ["openai", "ollama"]

[provider.keys]
anthropic = "sk-ant-api03-..."
openai = "sk-proj-..."
# google = "AIza..."

​
Option 2: Global API Key

Use api_key for single-provider setups: 
[provider]
primary = "anthropic"
model = "claude-sonnet-4-20250514"
api_key = "sk-ant-api03-..."

​
Option 3: Environment Variables

Set provider-specific env vars: 
export ANTHROPIC_API_KEY="sk-ant-..."
export OPENAI_API_KEY="sk-proj-..."
export DEEPSEEK_API_KEY="sk-..."
export GROQ_API_KEY="gsk-..."
export GOOGLE_API_KEY="AIza..."
Or use the global fallback: 
export ONECLAW_API_KEY="sk-ant-..."  # Used if no provider-specific key

​
Key Resolution Priority

  1. [provider.keys].<provider> in TOML
  2. api_key field in [provider] section
  3. ONECLAW_API_KEY environment variable
  4. Provider-specific env var (e.g., ANTHROPIC_API_KEY)

​
Ollama Configuration

Ollama runs locally and requires no API key: 
[provider]
primary = "ollama"
model = "llama3.2:3b"
max_tokens = 2048
temperature = 0.5

​
Custom Ollama Endpoint

For remote Ollama servers: 
[provider]
primary = "ollama"
ollama_endpoint = "http://192.168.1.100:11434"
ollama_model = "qwen2.5:3b"  # Override model for Ollama only
Note: ollama_model is separate from model to allow different models for primary vs. fallback.

​
Example Configurations

​
Edge Device (Raspberry Pi)

[provider]
primary = "anthropic"
model = "claude-haiku-4-5-20251001"  # Fast, cheap
max_tokens = 512                     # Limit response size
temperature = 0.2                    # Deterministic
fallback = ["ollama"]                # Local fallback

[provider.keys]
anthropic = "sk-ant-..."

​
Development / Testing

[provider]
primary = "ollama"                   # Free, local
model = "llama3.2:1b"                # Fast small model
max_tokens = 1024
temperature = 0.5

​
Production Multi-Cloud

[provider]
primary = "anthropic"
model = "claude-sonnet-4-20250514"
fallback = ["openai", "deepseek", "ollama"]
max_retries = 2
max_tokens = 2048
temperature = 0.3

[provider.keys]
anthropic = "sk-ant-..."
openai = "sk-proj-..."
deepseek = "sk-..."

​
Checking Provider Status

Use the providers command to see active provider configuration: 
providers
Output: 
Provider Chain:
  Primary:  anthropic/claude-sonnet-4-20250514 ✓
  Fallback: openai/gpt-4o ✓
  Fallback: ollama/llama3.2:3b ✓

Retries: 2 per provider
The ✓ indicates the provider is available (API key configured).

Build docs developers (and LLMs) love

Get started for freeTalk to us