Skip to main content
Beta
BetaServer tools are currently in beta. The API and behavior may change. This tool is experimental — when it graduates, the tool type is likely to be renamed (dropping the experimental__ prefix), which would be a breaking change for requests using the current name.
The openrouter:experimental__search_models server tool lets a model search and filter the OpenRouter model catalog. The model can look up models by name, capabilities, modalities, and other attributes — useful for agents that pick models dynamically, for example with the Subagent tool.

Quick Start

Configuration

The search models tool accepts an optional max_results parameter:
{
  "type": "openrouter:experimental__search_models",
  "parameters": {
    "max_results": 5
  }
}
ParameterTypeDefaultDescription
max_resultsinteger5Maximum number of models to return per call. Between 1 and 20

Call Arguments

The model generates the search arguments. All fields are optional — an empty call browses the full catalog:
FieldTypeDescription
querystringFree-text search matched against model names, slugs, and descriptions
input_modalitiesstring[]Filter by supported input modalities (text, image, file, audio, video). Returns models that support ALL specified modalities
output_modalitiesstring[]Filter by supported output modalities (e.g. text, image, embeddings, audio). Returns models that support ALL specified modalities
min_context_lengthintegerMinimum context length in tokens
seriesstringFilter by model series/family group (e.g. Claude, GPT, Gemini)

Response

The tool returns matching models with their key attributes:
{
  "models": [
    {
      "id": "anthropic/claude-sonnet-4.5",
      "name": "Anthropic: Claude Sonnet 4.5",
      "description": "...",
      "context_length": 1000000,
      "input_modalities": ["text", "image", "file"],
      "output_modalities": ["text"],
      "series": "Claude"
    }
  ],
  "total_results": 12,
  "showing": 1
}
total_results is the number of models that matched the filters; showing is how many were returned after applying max_results.

Pricing

There is currently no separate charge for the search models tool; you pay only for standard token usage.

Next Steps