# ConechoAI/openai-websearch-mcp [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/ConechoAI/openai-websearch-mcp/  
**GitHub Stars:** 93  
**npm Downloads (last month):** 49  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/conechoai-openai-websearch-mcp

## Description
This is a Python-based MCP server that provides OpenAI websearch built-in tool.

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

```json
"mcpServers": {
  "openai-websearch-mcp": {
    "command": "uvx",
    "args": ["--with"],
    "env": {
      "OPENAI_API_KEY": "",
      "OPENAI_DEFAULT_MODEL": ""
    }
  }
}
```

**Requires environment variables:** `OPENAI_API_KEY`, `OPENAI_DEFAULT_MODEL` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What ConechoAI/openai-websearch-mcp MCP server does

ConechoAI/openai-websearch-mcp MCP server makes OpenAI web search available to MCP-compatible assistants through one tool, `openai_web_search`. The tool accepts a search question or query and sends it to OpenAI's web search capability. It is intended for tasks where an assistant needs information that may have changed since its training data, such as current technology developments, research topics, or local events.

The server supports OpenAI models listed in the project documentation, including `gpt-4o`, `gpt-4o-mini`, `gpt-5`, `gpt-5-mini`, `gpt-5-nano`, `o3`, and `o4-mini`. Reasoning-capable models can be paired with a selected reasoning effort for faster searches or more extensive analysis.

## How it works

The server runs as a local MCP process and communicates with the client over the configured command. When the assistant calls `openai_web_search`, the server uses `OPENAI_API_KEY` to access OpenAI's API. The query is supplied through the required `input` parameter.

Optional parameters control the request:

- `model` selects the OpenAI model, with `gpt-5-mini` as the documented default.
- `reasoning_effort` accepts `low`, `medium`, `high`, or `minimal` where supported.
- `type` selects the web search API version and defaults to `web_search_preview`.
- `search_context_size` controls the amount of search context with `low`, `medium`, or `high`.
- `user_location` supplies location information for localized results.

The server applies reasoning settings only to models that support them, avoiding the unsupported-parameter error described for non-reasoning models.

## Setup and configuration

Install or run ConechoAI/openai-websearch-mcp MCP server with uvx:

```bash
uvx openai-websearch-mcp
```

It can also be installed from PyPI with `pip install openai-websearch-mcp` and started with `python -m openai_websearch_mcp`. The required environment variable is `OPENAI_API_KEY`. `OPENAI_DEFAULT_MODEL` optionally changes the default model from `gpt-5-mini`.

Claude Desktop and Cursor configurations use `uvx` as the command and `openai-websearch-mcp` as its argument. Claude Code can use the same MCP configuration as Claude Desktop. Local development is also documented with a virtual-environment Python executable and the module name.

## Tools and capabilities

ConechoAI/openai-websearch-mcp MCP server provides the `openai_web_search` tool. Its capabilities include:

- Current web searches through OpenAI's Web Search API.
- Model selection across standard and reasoning models documented by the project.
- Configurable reasoning effort for compatible models.
- Adjustable search context size.
- Optional location-aware search results.
- Fast multi-round searches or deeper research workflows, depending on model and reasoning settings.

## Limitations and notes

An OpenAI API key is required, and API usage may incur OpenAI charges. The server is a local wrapper around OpenAI's web search functionality; the documentation does not describe an independent search index or offline mode. Reasoning parameters should not be used with models that do not support them, although the server handles this compatibility case automatically. The documented integrations include Claude Desktop, Cursor, and Claude Code; support for other clients is not established by the provided material.

_Full upstream README: https://allmcps.com/mcp/conechoai-openai-websearch-mcp/readme_