OpenRouter Provider

Access 200+ models through OpenRouter’s unified API

OpenRouter provides access to models from multiple providers through a single API, with automatic fallbacks and load balancing.

Overview

OpenRouter aggregates models from:

• OpenAI (GPT-4, GPT-3.5)
• Anthropic (Claude)
• Google (Gemini)
• Meta (Llama)
• Mistral, Cohere, and more

Supports:

• 200+ models
• Automatic failover
• Usage-based pricing
• No commitment

Configuration

Environment Variable

export OPENROUTER_API_KEY="sk-or-v1-..."

Config File

[agent]
provider = "openrouter"
model = "anthropic/claude-3.5-sonnet"  # Model name from OpenRouter catalog

[providers.openrouter]
api_key = "sk-or-v1-..."  # Optional: defaults to OPENROUTER_API_KEY
base_url = "https://openrouter.ai/api/v1"  # Optional: custom endpoint

Selecting Models

OpenRouter uses slash-separated model IDs:

# Anthropic models
zeroclaw agent --model anthropic/claude-3.5-sonnet
zeroclaw agent --model anthropic/claude-3-opus

# OpenAI models
zeroclaw agent --model openai/gpt-4-turbo
zeroclaw agent --model openai/gpt-3.5-turbo

# Open source models
zeroclaw agent --model meta-llama/llama-3.1-70b-instruct
zeroclaw agent --model mistralai/mixtral-8x7b-instruct

Browse available models at: https://openrouter.ai/models

Features

Tool Calling

Native tool calling for compatible models:

[providers.openrouter]
tool_calling = "native"  # Uses model's native function calling

Models with tool support:

• anthropic/claude-3.5-sonnet
• openai/gpt-4-turbo
• google/gemini-pro

Automatic Fallbacks

OpenRouter automatically retries with alternative models if primary fails:

[providers.openrouter]
fallback_models = [
    "anthropic/claude-3.5-sonnet",
    "openai/gpt-4-turbo",
    "google/gemini-pro"
]

Cost Tracking

Monitor spending:

zeroclaw cost summary --provider openrouter

Request Format

OpenRouter uses OpenAI-compatible request format:

{
  "model": "anthropic/claude-3.5-sonnet",
  "messages": [
    {"role": "user", "content": "Hello!"}
  ],
  "temperature": 0.7,
  "max_tokens": 1000
}

Rate Limits

Varies by model and plan:

• Free tier: 10 requests/minute
• Pay-as-you-go: 60 requests/minute
• Higher tiers: Contact OpenRouter

Pricing

Pay only for what you use. Prices vary by model:

Model	Input (per 1M tokens)	Output (per 1M tokens)
Claude 3.5 Sonnet	$3.00	$15.00
GPT-4 Turbo	$10.00	$30.00
Llama 3.1 70B	$0.59	$0.79

Check current pricing at: https://openrouter.ai/models

Troubleshooting

'Invalid API key' error

Solution:

Get your API key from https://openrouter.ai/keys

export OPENROUTER_API_KEY="sk-or-v1-..."
zeroclaw status

'Model not found' error

Solution:

Check model ID format (should be provider/model-name):

# Correct
zeroclaw agent --model anthropic/claude-3.5-sonnet

# Wrong
zeroclaw agent --model claude-3.5-sonnet

Rate limit errors

Solution:

Add rate limiting:

[security]
max_actions_per_hour = 50
rate_limit_per_minute = 10

Example Usage

# Use Claude via OpenRouter
export OPENROUTER_API_KEY="sk-or-v1-..."
zeroclaw agent --provider openrouter --model anthropic/claude-3.5-sonnet

# Try multiple models with fallback
zeroclaw agent -p openrouter -m "openai/gpt-4-turbo,anthropic/claude-3-opus"

Related

• Custom Provider Setup
• Provider Trait
• Configuration