OpenRouter Setup

OpenRouter provides access to multiple AI providers through a single API, giving you flexibility to switch between models like GPT-4, Claude, Gemini, and others.

Features

Unified API: Access multiple AI providers with one API key
Model Flexibility: Switch between models without code changes
Competitive Pricing: Pay-per-use with transparent pricing
No Rate Limits: Less restrictive than individual provider limits

Setup

Get API Key

Visit OpenRouter and sign up for an account.

Navigate to Keys section
Create a new API key
Copy the key (starts with sk-or-v1-)

Configure Environment

Add your OpenRouter settings to .env:

OPENROUTER_API_KEY=sk-or-v1-your_api_key_here
OPENROUTER_MODEL=google/gemini-2.5-flash

Keep your API key secure. Never commit it to version control.

Enable OpenRouter in Code

OpenRouter is enabled when you pass the useOpenRouter parameter:

const llmHelper = new LLMHelper(
  undefined,           // apiKey (not needed)
  false,              // useOllama
  undefined,          // ollamaModel
  undefined,          // ollamaUrl
  true,               // useOpenRouter ✓
  process.env.OPENROUTER_API_KEY,
  process.env.OPENROUTER_MODEL
)

Verify Configuration

Start Cluely and check for:

[LLMHelper] Using OpenRouter with model: google/gemini-2.5-flash

Configuration

Environment Variables

Variable	Required	Description	Default
`OPENROUTER_API_KEY`	Yes	Your OpenRouter API key	-
`OPENROUTER_MODEL`	No	Model to use	`google/gemini-2.5-flash`

Supported Models

Cluely is tested with these models, but OpenRouter supports many more:

Gemini Models
GPT Models
Claude Models
Open Source Models

OPENROUTER_MODEL=google/gemini-2.5-flash
OPENROUTER_MODEL=google/gemini-pro
OPENROUTER_MODEL=google/gemini-pro-vision

Pros: Fast, high-quality, vision support on some models Pricing: Low cost per token

OPENROUTER_MODEL=openai/gpt-4
OPENROUTER_MODEL=openai/gpt-4-turbo
OPENROUTER_MODEL=openai/gpt-3.5-turbo

Pros: Excellent reasoning, wide knowledge base Pricing: Moderate to high cost

OPENROUTER_MODEL=anthropic/claude-3-opus
OPENROUTER_MODEL=anthropic/claude-3-sonnet
OPENROUTER_MODEL=anthropic/claude-3-haiku

Pros: Strong reasoning, good for long conversations Pricing: Moderate cost

OPENROUTER_MODEL=meta-llama/llama-3-70b
OPENROUTER_MODEL=mistralai/mistral-large
OPENROUTER_MODEL=deepseek/deepseek-chat

Pros: Lower cost, good performance Pricing: Very low cost

See the full list of models at OpenRouter Models

API Implementation

Request Configuration

Cluely sends requests to OpenRouter with this configuration:

// OpenRouter API call (source/electron/LLMHelper.ts:242-261)
const response = await fetch("https://openrouter.ai/api/v1/chat/completions", {
  method: 'POST',
  headers: {
    'Authorization': `Bearer ${this.openRouterApiKey}`,
    'Content-Type': 'application/json',
    'HTTP-Referer': 'https://cluely.ai',
    'X-Title': 'Cluely'
  },
  body: JSON.stringify({
    model: this.openRouterModel,
    messages: [
      {
        role: "user",
        content: prompt
      }
    ],
    temperature: 0.7,
    max_tokens: 4096
  }),
})

Endpoint

POST https://openrouter.ai/api/v1/chat/completions

Headers

Header	Value	Purpose
`Authorization`	`Bearer {API_KEY}`	Authentication
`Content-Type`	`application/json`	Request format
`HTTP-Referer`	`https://cluely.ai`	App identification
`X-Title`	`Cluely`	App name

Model Selection

Default Model

Cluely uses google/gemini-2.5-flash as the default OpenRouter model:

// Default OpenRouter model (source/electron/LLMHelper.ts:33)
private openRouterModel: string = "google/gemini-2.5-flash"

Changing Models

OPENROUTER_MODEL=anthropic/claude-3-opus

Limitations with OpenRouter

OpenRouter has some limitations in Cluely’s current implementation:

Vision/Image Analysis

OpenRouter cannot directly analyze images in the current implementation:

// Guidance-only approach (source/electron/LLMHelper.ts:381-392)
if (this.useOpenRouter) {
  const imageCount = imagePaths.length;
  const prompt = `I have ${imageCount} screenshot(s) that I need help analyzing...`
  // Provides guidance instead of actual image analysis
}

Workaround: Switch to Gemini for screenshot analysis features.

Audio Processing

Audio features are not supported with OpenRouter. Cluely uses Gemini for all voice functionality.

Switching to OpenRouter

At Runtime

Switch from another provider to OpenRouter:

// Switch to OpenRouter
await llmHelper.switchToOpenRouter(
  'sk-or-v1-your_api_key',
  'google/gemini-2.5-flash'
)

// Verify current provider
const provider = llmHelper.getCurrentProvider()
console.log(provider)  // "openrouter"

const model = llmHelper.getCurrentModel()
console.log(model)  // "google/gemini-2.5-flash"

Check Connection

Test your OpenRouter configuration:

const result = await llmHelper.testConnection()

if (result.success) {
  console.log('OpenRouter connected successfully')
} else {
  console.error('Connection failed:', result.error)
}

Cost Management

Monitor Usage

Track your OpenRouter usage:

Visit OpenRouter Dashboard
View usage by model and date
Set spending limits to prevent overages

Cost Optimization

Choose Appropriate Models

For simple tasks: Use google/gemini-2.5-flash or openai/gpt-3.5-turbo
For complex reasoning: Use openai/gpt-4 or anthropic/claude-3-opus
For cost efficiency: Use open-source models like meta-llama/llama-3-70b

Optimize Request Parameters

// Reduce max_tokens for shorter responses
max_tokens: 2048  // instead of 4096

// Lower temperature for more focused responses
temperature: 0.5  // instead of 0.7

Use Rate Limiting

Implement client-side rate limiting to control costs:

Limit requests per minute
Cache common responses
Debounce user input

Troubleshooting

API Key Not Working

Error: OpenRouter API key is not configuredSolutions:

Verify OPENROUTER_API_KEY is set in .env
Check API key starts with sk-or-v1-
Ensure no extra spaces or quotes in .env
Restart Cluely after updating .env

Invalid Model

Error: Model not found or similarSolutions:

Check model name is correct (case-sensitive)
Verify model exists: OpenRouter Models
Use full model path: google/gemini-2.5-flash not just gemini-2.5-flash

Rate Limiting

Error: 429 Too Many RequestsSolutions:

Wait before retrying (OpenRouter handles rate limits per model)
Switch to a different model with higher limits
Upgrade your OpenRouter plan

Insufficient Credits

Error: Payment or credit errorsSolutions:

Add credits in OpenRouter dashboard
Update payment method
Check spending limits aren’t exceeded

Best Practices

Start with default model: google/gemini-2.5-flash offers good balance of cost and quality
Monitor usage: Check OpenRouter dashboard regularly
Test before production: Verify model behavior matches expectations
Set spending limits: Prevent unexpected charges
Use appropriate models: Don’t use expensive models for simple tasks

Hybrid Setup

Combine OpenRouter with other providers:

// Use OpenRouter for text chat
const llmHelper = new LLMHelper(
  process.env.GEMINI_API_KEY,      // Still needed for voice
  false,
  undefined,
  undefined,
  true,                             // Enable OpenRouter
  process.env.OPENROUTER_API_KEY,
  'anthropic/claude-3-opus'
)

// Gemini handles voice automatically
// OpenRouter handles text chat

Get Started

Core Features

AI Providers

Guides

OpenRouter Setup

Features

Setup

Configuration

Environment Variables

Supported Models

API Implementation

Request Configuration

Endpoint

Headers

Model Selection

Default Model

Changing Models

Limitations with OpenRouter

Vision/Image Analysis

Audio Processing

Switching to OpenRouter

At Runtime

Check Connection

Cost Management

Monitor Usage

Cost Optimization

Troubleshooting

Best Practices

Hybrid Setup

Next Steps

Model Comparison

K2 Think Setup

Build docs developers (and LLMs) love

Get Started

Core Features

AI Providers

Guides

​Features

​Setup

​Configuration

​Environment Variables

​Supported Models

​API Implementation

​Request Configuration

​Endpoint

​Headers

​Model Selection

​Default Model

​Changing Models

​Limitations with OpenRouter

​Vision/Image Analysis

​Audio Processing

​Switching to OpenRouter

​At Runtime

​Check Connection

​Cost Management

​Monitor Usage

​Cost Optimization

​Troubleshooting

​Best Practices

​Hybrid Setup

​Next Steps

Model Comparison

K2 Think Setup

Build docs developers (and LLMs) love

Features

Setup

Configuration

Environment Variables

Supported Models

API Implementation

Request Configuration

Endpoint

Headers

Model Selection

Default Model

Changing Models

Limitations with OpenRouter

Vision/Image Analysis

Audio Processing

Switching to OpenRouter

At Runtime

Check Connection

Cost Management

Monitor Usage

Cost Optimization

Troubleshooting

Best Practices

Hybrid Setup

Next Steps