assafelovic/gpt-researcher · critical · ValueError

MCPRetriever requires a researcher instance with cfg attribu

Error message

MCPRetriever requires a researcher instance with cfg attribute containing LLM configuration

What it means

ValueError raised by MCPRetriever._get_config() when the retriever was constructed without a researcher instance (or with one lacking a cfg attribute). The retriever needs the researcher's config to obtain LLM settings for its two-stage search, so it treats a missing config as a critical, non-recoverable error.

Source

Thrown at gpt_researcher/retrievers/mcp/retriever.py:114

            List[Dict[str, Any]]: List of MCP server configurations.
        """
        if self.researcher and hasattr(self.researcher, 'mcp_configs'):
            return self.researcher.mcp_configs or []
        return []

    def _get_config(self):
        """
        Get configuration from the researcher instance.
        
        Returns:
            Config: Configuration object with LLM settings.
        """
        if self.researcher and hasattr(self.researcher, 'cfg'):
            return self.researcher.cfg
        
        # If no config available, this is a critical error
        logger.error("No config found in researcher instance. MCPRetriever requires a researcher instance with cfg attribute.")
        raise ValueError("MCPRetriever requires a researcher instance with cfg attribute containing LLM configuration")

    async def search_async(self, max_results: int = 10) -> List[Dict[str, str]]:
        """
        Perform an async search using MCP tools with intelligent two-stage approach.
        
        Args:
            max_results: Maximum number of results to return.
            
        Returns:
            List[Dict[str, str]]: The search results.
        """
        # Check if we have any server configurations
        if not self.mcp_configs:
            error_msg = "No MCP server configurations available. Please provide mcp_configs parameter to GPTResearcher."
            logger.error(error_msg)
            await self.streamer.stream_error("MCP retriever cannot proceed without server configurations.")
            return []  # Return empty instead of raising to allow research to continue

View on GitHub (pinned to 6f998577d5)

Solutions

  1. Pass the researcher instance: MCPRetriever(query, researcher=self.researcher).
  2. If testing, use a simple stub: types.SimpleNamespace(cfg=your_cfg) as the researcher.
  3. Ensure you construct retrievers through the standard GPTResearcher flow so cfg is populated.

Example fix

# before
retriever = MCPRetriever(query="...")

# after
retriever = MCPRetriever(query="...", researcher=researcher_instance)
# tests: MCPRetriever(q, researcher=SimpleNamespace(cfg=test_cfg))
Defensive patterns

Strategy: type-guard

Validate before calling

assert researcher is not None and hasattr(researcher, "cfg"), "researcher with cfg required"

Type guard

def has_cfg(r) -> bool:
    return r is not None and hasattr(r, "cfg") and r.cfg is not None

Try / catch

try:
    retriever = MCPRetriever(query, researcher=researcher)
except ValueError as e:
    if "cfg attribute" in str(e):
        raise TypeError("Pass the GPTResearcher instance") from e
    raise

Prevention

When it happens

Trigger: Creating MCPRetriever(query) without passing the researcher, or passing a mock/stand-in object that has no cfg attribute; __init__ calls _get_config() immediately.

Common situations: Using the retriever standalone in tests or scripts, refactoring that drops the researcher argument, or constructing it with a None researcher expecting a default config path.

Related errors

AI-assisted analysis of assafelovic/gpt-researcher@6f998577d5 (2026-08-28). Data as JSON: /api/errors/3de076eab7f482c0. Report an issue: GitHub.