# reading-plus-ai/mcp-server-deep-research [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/reading-plus-ai/mcp-server-deep-research  
**GitHub Stars:** 215  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/reading-plus-ai-mcp-server-deep-research

## Description
MCP server providing OpenAI/Perplexity-like autonomous deep research, structured query elaboration, and concise reporting.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "mcp-server-deep-research": {
    "command": "uvx",
    "args": ["--directory","/Users/username/repos/mcp-server-application/mcp-server-deep-research","run","mcp-server-deep-research"]
  }
}
```

## Documentation

## What reading-plus-ai/mcp-server-deep-research MCP server does

The reading-plus-ai/mcp-server-deep-research MCP server provides a guided workflow for researching complex questions. Instead of treating the initial prompt as a single search query, it first clarifies the subject, identifies important terms, and establishes the intended scope. It then breaks the topic into focused subquestions so different aspects can be investigated separately.

The workflow is intended to produce a readable research report rather than a list of search results. Reports are structured as artifacts, include citations for the sources used, and present conclusions with supporting evidence. The documented process also aims to include relevant and authoritative sources and to represent different perspectives where applicable.

## How it works

The main entry point described in the repository is the `deep-research` prompt. A user supplies a research question through the MCP workflow, and the prompt guides the subsequent stages:

- clarify the question and define its boundaries;
- generate subquestions for systematic coverage;
- search the web for each subquestion using Claude’s built-in web search;
- assess source quality and relevance;
- combine findings from multiple sources; and
- format the results into a cited report artifact.

The reading-plus-ai/mcp-server-deep-research MCP server therefore fits tasks where the research plan should be generated from the question itself. The README does not describe a separate search API credential, an external search provider, or a set of individually named research tools.

## Setup and configuration

The README documents Claude Desktop as the client used to try the project. On macOS, the stated setup command is `python setup.py`. After setup, the user selects the `deep-research` prompt template and provides a research question.

For a published installation, the Claude Desktop configuration uses `uvx` with the package name `mcp-server-deep-research`:

```json
{
  "mcpServers": {
    "mcp-server-deep-research": {
      "command": "uvx",
      "args": ["mcp-server-deep-research"]
    }
  }
}
```

Claude Desktop stores its configuration at `~/Library/Application Support/Claude/claude_desktop_config.json` on macOS and `%APPDATA%/Claude/claude_desktop_config.json` on Windows. For unpublished development copies, the README shows a `uv` configuration that points to a local project directory; that example contains a machine-specific path and is not a directly runnable install command.

## Tools and capabilities

The reading-plus-ai/mcp-server-deep-research MCP server documents one prompt component: `deep-research`. Its supported workflow capabilities include question elaboration, subquestion generation, web search through Claude, source analysis, synthesis, citation, and report generation. The report output is described as a formatted artifact with a balanced, evidence-based presentation.

The project can also be developed and packaged with `uv`. The documented development commands synchronize dependencies with `uv sync`, build distributions with `uv build`, and publish them with `uv publish`.

## Limitations and notes

The repository material describes a Claude-centered workflow and specifically instructs users to use Claude Desktop. It does not establish compatibility with other MCP clients. Web searching is described as using Claude’s built-in capability, so the README does not provide configuration for an independent search service or a search API key.

Only the `deep-research` prompt is listed in the components section. The available material does not specify adjustable research-depth settings, database storage, scheduled jobs, custom source connectors, or additional standalone tools. The setup instructions also differ between a local unpublished checkout and the published `uvx` package, so the appropriate configuration depends on whether the project is being developed locally or installed as a published server.

_Full upstream README: https://allmcps.com/mcp/reading-plus-ai-mcp-server-deep-research/readme_