Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.parallel.ai/llms.txt

Use this file to discover all available pages before exploring further.

Parallel is available as a web search engine on OpenRouter, enabling any model to use Parallel-powered web search results. Select Parallel as the engine in the OpenRouter web_search server tool to ground AI responses with real-time web data. Read OpenRouter’s official documentation here.

​
How it works

OpenRouter’s web_search server tool lets the model decide when and how often to search the web. When you select Parallel as the engine, OpenRouter routes the search request through Parallel’s Search API and returns the results as annotations on the model response. This works with any model on OpenRouter, regardless of the provider.
OpenRouter’s older web plugin is deprecated. New integrations should use the openrouter:web_search server tool shown below.

​
Quick start

Set the engine to "parallel" in the openrouter:web_search server tool: 
{
  "model": "openai/gpt-5.5",
  "messages": [
    {
      "role": "user",
      "content": "What are the latest developments in quantum computing?"
    }
  ],
  "tools": [
    {
      "type": "openrouter:web_search",
      "parameters": {
        "engine": "parallel"
      }
    }
  ]
}

​
Configuration options

Customize the server tool with additional parameters: 
{
  "model": "openai/gpt-5.5",
  "tools": [
    {
      "type": "openrouter:web_search",
      "parameters": {
        "engine": "parallel",
        "max_results": 5,
        "max_total_results": 20,
        "search_context_size": "medium",
        "allowed_domains": ["arxiv.org", "nature.com"]
      }
    }
  ]
}
ParameterTypeDefaultDescription
enginestring—Set to "parallel" to use Parallel search
max_resultsinteger5Results per search call (range 1–25)
max_total_resultsinteger—Cap on cumulative results across multiple searches in a single response
search_context_sizestringmediumlow, medium, or high. For Parallel, controls the total characters returned across all results
allowed_domainsstring[]—Only include results from these domains
excluded_domainsstring[]—Exclude results from these domains
allowed_domains and excluded_domains are mutually exclusive when using Parallel as the engine — you cannot use both in the same request.

​
Domain filtering

Restrict which domains appear in search results: 
{
  "model": "anthropic/claude-opus-4.7",
  "tools": [
    {
      "type": "openrouter:web_search",
      "parameters": {
        "engine": "parallel",
        "allowed_domains": ["arxiv.org", ".github.io"]
      }
    }
  ]
}
Both fields accept plain domains (e.g., parallel.ai) or bare domain extensions starting with a period (e.g., .edu, .gov).

​
Parsing search results

Web search results are returned as annotations in the OpenAI Chat Completion format: 
{
  "message": {
    "role": "assistant",
    "content": "Here's what I found: ...",
    "annotations": [
      {
        "type": "url_citation",
        "url_citation": {
          "url": "https://www.example.com/article",
          "title": "Article Title",
          "content": "Relevant content from the page",
          "start_index": 100,
          "end_index": 200
        }
      }
    ]
  }
}

​
Pricing

When using Parallel as the engine on OpenRouter, pricing has two components:
  1. Parallel Search API costs — see Parallel pricing for current rates
  2. LLM usage costs — charged by OpenRouter for the additional prompt tokens from search results