OpenRouter LLM plugin guide

Available in

Python

Overview

This plugin allows you to use OpenRouter as an LLM provider for your voice agents. OpenRouter provides access to hundreds of models from multiple providers through a unified API, with automatic fallback support and intelligent routing.

Usage

Install the OpenAI plugin to add OpenRouter support:

uv add "livekit-agents[openai]~=1.2"

Authentication

The OpenRouter plugin requires an OpenRouter API key.

Set OPENROUTER_API_KEY in your .env file.

Create an OpenRouter LLM using the with_openrouter method:

from livekit.plugins import openai

session = AgentSession(
    llm=openai.LLM.with_openrouter(model="anthropic/claude-sonnet-4.5"),
    # ... tts, stt, vad, turn_detection, etc.
)

Parameters

This section describes some of the available parameters. See the usage examples below and the plugin reference links in the Additional resources section for a complete list of all available parameters.

modelstringOptionalDefault: openrouter/auto

Model to use. Can be "openrouter/auto" to let OpenRouter choose, or specify a specific model like "anthropic/claude-sonnet-4.5". For a list of available models, see OpenRouter models.

site_urlstringOptional

Your site URL for analytics and ranking on OpenRouter. This is sent as the HTTP-Referer header.

app_namestringOptional

Your app name for analytics on OpenRouter. This is sent as the X-Title header.

fallback_modelslist[string]Optional

List of fallback models to use if the primary model is unavailable. Example: fallback_models=["anthropic/claude-sonnet-4", "openai/gpt-5-mini"].

providerdictOptional

Provider routing preferences for fine-grained control over model selection. Can include:

order: List of preferred providers in order
allow_fallbacks: Whether to allow fallback to other providers
require_parameters: Whether to require specific parameters
data_collection: Data collection preference, either "allow" or "deny"
only: List of providers to exclusively use
ignore: List of providers to exclude
quantizations: List of accepted quantization levels
sort: Sort providers by "price", "throughput", or "latency"
max_price: Maximum price per token

Refer to the OpenRouter documentation for more information.

pluginslist[OpenRouterWebPlugin]Optional

List of OpenRouter plugins to enable. Currently supports web search plugin with configuration for max results and search prompts.

Examples

The following examples demonstrate usage of various OpenRouter parameters.

Configure multiple fallback models to use if the primary model is unavailable:

from livekit.plugins import openai

llm = openai.LLM.with_openrouter(
    model="openai/gpt-4o",
    fallback_models=[
        "anthropic/claude-sonnet-4",
        "openai/gpt-5-mini",
    ],
)

Control which providers are used for model inference:

from livekit.plugins import openai

llm = openai.LLM.with_openrouter(
    model="deepseek/deepseek-chat-v3.1",
    provider={
        "order": ["novita/fp8", "gmicloud/fp8", "google-vertex"],
        "allow_fallbacks": True,
        "sort": "latency",
    },
)

Enable OpenRouter's web search capabilities:

from livekit.plugins import openai

llm = openai.LLM.with_openrouter(
    model="google/gemini-2.5-flash-preview-09-2025",
    plugins=[
        openai.OpenRouterWebPlugin(
            max_results=5,
            search_prompt="Search for relevant information",
        )
    ],
)

Include site and app information for OpenRouter analytics:

from livekit.plugins import openai

llm = openai.LLM.with_openrouter(
    model="openrouter/auto",
    site_url="https://myapp.com",
    app_name="My Voice Agent",
)

Additional resources

The following links provide more information about the OpenRouter integration.

Python plugin

Reference GitHub PyPI

GitHub repo

View the source or contribute to the LiveKit OpenAI LLM plugin.

OpenRouter docs

OpenRouter API documentation and model list.

Voice AI quickstart

Get started with LiveKit Agents and OpenRouter.