Add MCP Server for Agent Interoperability

Is your feature request related to a problem?

Thanks for this powerful, standalone tool for performing detailed research. However, because it's run from the command line, it's difficult to integrate as a "tool" into larger AI agent frameworks or multi-agent systems. A more sophisticated agent has no standardized way to discover what this tool does or how to call it.

Describe the solution you'd like

I propose adding a simple MCP (Model Context Protocol) server to this project.

This would expose the core research functionality as a callable tool over an API. When another agent queries the MCP endpoint, it would receive a structured response describing how to use the research function.

For example, the server could be started with npm run server and would expose a context like this:

{
  "functions": [
    {
      "name": "perform_deep_research",
      "description": "Performs iterative, deep research on a given topic by combining search, web scraping, and LLMs to produce a detailed report.",
      "parameters": {
        "type": "object",
        "properties": {
          "topic": {
            "type": "string",
            "description": "The central topic or question to research."
          },
          "depth": {
            "type": "integer",
            "description": "How many recursive levels of research to perform (e.g., 2)."
          },
          "breadth": {
            "type": "integer",
            "description": "How many search results to explore at each level (e.g., 4)."
          }
        },
        "required": ["topic"]
      }
    }
  ]
}

Why is this a good idea?

• Makes it a Composable Tool: This immediately turns open-deep-research from a script into a building block that can be used by other AI agents.
• Standardized Integration: MCP is becoming a standard for agent-to-tool communication. Supporting it makes this project instantly compatible with a growing ecosystem.
• Increases Utility: It allows this project's powerful research capability to be called programmatically, unlocking its use in autonomous systems, chatbots, and more complex workflows.