OpenRouter

Access 200+ models from multiple providers (OpenAI, Anthropic, Google, Meta, Mistral, and more) through a single API key and unified interface.

Prerequisites

• An OpenRouter API key from openrouter.ai

Quick Setup

1. Get API Key

1. Sign up at openrouter.ai
2. Go to Keys in your dashboard
3. Click Create Key and copy it (starts with sk-or-)

2. Configure

[default]
provider = "openrouter"
model = "anthropic/claude-sonnet-4"

[providers.openrouter]
api_key = "${OPENROUTER_API_KEY}"

Or set the environment variable:

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

3. Verify

prx doctor models

Available Models

OpenRouter provides access to hundreds of models. Some popular options:

Model	Provider	Context	Vision	Tool Use	Notes
anthropic/claude-sonnet-4	Anthropic	200K	Yes	Yes	Claude Sonnet 4
anthropic/claude-opus-4	Anthropic	200K	Yes	Yes	Claude Opus 4
openai/gpt-4o	OpenAI	128K	Yes	Yes	GPT-4o
openai/o3	OpenAI	128K	Yes	Yes	Reasoning model
google/gemini-2.5-pro	Google	1M	Yes	Yes	Gemini Pro
google/gemini-2.5-flash	Google	1M	Yes	Yes	Gemini Flash
meta-llama/llama-3.1-405b-instruct	Meta	128K	No	Yes	Largest open model
deepseek/deepseek-chat	DeepSeek	128K	No	Yes	DeepSeek V3
mistralai/mistral-large	Mistral	128K	No	Yes	Mistral Large
x-ai/grok-2	xAI	128K	No	Yes	Grok 2

Browse the full model list at openrouter.ai/models.

Configuration Reference

Field	Type	Default	Description
api_key	string	required	OpenRouter API key (sk-or-...)
model	string	required	Model ID in provider/model format

Features

Unified Multi-Provider Access

With a single OpenRouter API key, you can access models from OpenAI, Anthropic, Google, Meta, Mistral, Cohere, and many more. This eliminates the need to manage multiple API keys.

OpenAI-Compatible API

OpenRouter exposes an OpenAI-compatible Chat Completions API at https://openrouter.ai/api/v1/chat/completions. PRX sends requests with:

• Authorization: Bearer <key> for authentication
• HTTP-Referer: https://github.com/theonlyhennygod/openprx for app identification
• X-Title: OpenPRX for app name attribution

Native Tool Calling

Tools are sent in OpenAI's native function-calling format. The provider supports tool_choice: "auto" and properly handles structured tool call responses including tool_call_id mapping for multi-turn tool interactions.

Multi-Turn Conversation History

Full conversation history is preserved with proper structured formatting:

• Assistant messages with tool calls are serialized with tool_calls arrays
• Tool result messages include tool_call_id references
• System, user, and assistant messages are passed through directly

Connection Warmup

On startup, PRX sends a lightweight request to https://openrouter.ai/api/v1/auth/key to verify the API key and establish TLS/HTTP2 connection pooling.

Model Routing

OpenRouter supports model routing and fallback at the API level. You can also use PRX's built-in fallback_providers for client-side fallback:

[default]
provider = "openrouter"
model = "anthropic/claude-sonnet-4"

[reliability]
fallback_providers = ["openai"]

Default Provider

OpenRouter is PRX's default provider. If no provider is specified in your configuration, PRX defaults to OpenRouter.

Troubleshooting

"OpenRouter API key not set"

Set the OPENROUTER_API_KEY environment variable or add api_key under [providers.openrouter] in your config.toml. You can also run prx onboard for interactive setup.

402 Payment Required

Your OpenRouter account has insufficient credits. Add credits at openrouter.ai/credits.

Model-specific errors

Different models on OpenRouter have different capabilities and rate limits. If a specific model returns errors:

• Check if the model supports tool calling (not all do)
• Verify the model is not deprecated on OpenRouter
• Try a different model variant

Slow responses

OpenRouter routes to the underlying provider. Response time depends on:

• The model provider's current load
• Your geographic distance from the provider
• The model size and context length

Consider using fallback_providers to fail over to a direct provider connection if OpenRouter is slow.

Rate limiting

OpenRouter has its own rate limits in addition to underlying provider limits. If rate-limited:

• Check your usage at openrouter.ai/usage
• Upgrade your plan for higher limits
• Use PRX's reliable provider wrapper for automatic retry with backoff