search_available_models

Find AI models for image generation by searching with descriptive queries and optional style filters through the Replicate API.

Instructions

Search for available models matching the query.

    Args:
        query: Search query describing the desired model
        style: Optional style to filter by

    Returns:
        List of matching models with scores

Input Schema

TableJSON Schema

Name	Required	Description	Default
`query`	Yes
`style`	No

Implementation Reference

src/mcp_server_replicate/server.py:927-971 (handler)

Handler function for the search_available_models tool. Searches Replicate models using the client, scores them based on popularity, featured status, version stability, tags matching style and image generation, sorts by score, and returns a ModelList.

async def search_available_models(
    query: str,
    style: str | None = None,
) -> ModelList:
    """Search for available models matching the query.

    Args:
        query: Search query describing the desired model
        style: Optional style to filter by

    Returns:
        List of matching models with scores
    """
    search_query = query
    if style:
        search_query = f"{style} style {search_query}"

    async with ReplicateClient(api_token=os.getenv("REPLICATE_API_TOKEN")) as client:
        result = await client.search_models(search_query)
        models = [Model(**model) for model in result["models"]]

        # Score models but don't auto-select
        scored_models = []
        for model in models:
            score = 0
            run_count = getattr(model, "run_count", 0) or 0
            score += min(50, (run_count / 1000) * 50)
            if getattr(model, "featured", False):
                score += 20
            if model.latest_version:
                score += 10
            tags = getattr(model, "tags", [])
            if style and any(style.lower() in tag.lower() for tag in tags):
                score += 15
            if "image" in tags or "text-to-image" in tags:
                score += 15
            scored_models.append((model, score))

        # Sort by score but return all for user selection
        scored_models.sort(key=lambda x: x[1], reverse=True)
        return ModelList(
            models=[m[0] for m in scored_models],
            next_cursor=result.get("next_cursor"),
            total_count=result.get("total_count"),
        )

src/mcp_server_replicate/server.py:927-927 (registration)
The @mcp.tool() decorator registers the search_available_models function as an MCP tool with the function name.
```
async def search_available_models(
```

src/mcp_server_replicate/models/model.py:25-30 (schema)

Pydantic model defining the output schema for search_available_models, used as return type.

github_url: Optional[str] = Field(None, description="URL to model's GitHub repository")
paper_url: Optional[str] = Field(None, description="URL to model's research paper")
license_url: Optional[str] = Field(None, description="URL to model's license")
run_count: Optional[int] = Field(None, description="Number of times this model has been run")
cover_image_url: Optional[str] = Field(None, description="URL to model's cover image")
latest_version: Optional[ModelVersion] = Field(None, description="Latest version of the model")

MCP Server Replicate