The OpenAI WebSearch MCP Server enables AI assistants to perform intelligent web searches with advanced reasoning capabilities and real-time information access.
Core Capabilities:
Intelligent web search using OpenAI's reasoning models for up-to-date information retrieval
Multi-model support including GPT-4o, GPT-4o-mini, GPT-5, GPT-5-mini, GPT-5-nano, o3, and o4-mini
Configurable reasoning effort with low, medium, high, or minimal levels for optimal performance
Multiple search modes - fast iterations with lightweight models or comprehensive deep research
Localized search results based on user location (city, timezone, region, country)
Adjustable search context with low, medium, or high settings
Multiple API versions including web_search_preview and web_search_preview_2025_03_11
Integration & Configuration:
Seamless integration with AI assistants like Claude Desktop, Cursor, and Claude Code via MCP protocol
Easy configuration through environment variables for API keys and default models
Flexible parameters for customizing each search query including input type and model selection
Provides access to OpenAI's websearch tool to query for current information from the web
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@OpenAI WebSearch MCP Server search for the latest AI breakthroughs using gpt-5 with high reasoning effort"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An advanced MCP server that provides intelligent web search capabilities using OpenAI's reasoning models. Perfect for AI assistants that need up-to-date information with smart reasoning capabilities.
🧠 Reasoning Model Support: Full compatibility with OpenAI's latest reasoning models (gpt-5, gpt-5-mini, gpt-5-nano, o3, o4-mini)
⚡ Smart Effort Control: Intelligent reasoning_effort defaults based on use case
🔄 Multi-Mode Search: Fast iterations with gpt-5-mini or deep research with gpt-5
🌍 Localized Results: Support for location-based search customization
📝 Rich Descriptions: Complete parameter documentation for easy integration
🔧 Flexible Configuration: Environment variable support for easy deployment
Related MCP server: OneSearch MCP Server
OPENAI_API_KEY=sk-xxxx uvx --with openai-websearch-mcp openai-websearch-mcp-installReplace sk-xxxx with your OpenAI API key from the OpenAI Platform.
Add to your claude_desktop_config.json:
{
"mcpServers": {
"openai-websearch-mcp": {
"command": "uvx",
"args": ["openai-websearch-mcp"],
"env": {
"OPENAI_API_KEY": "your-api-key-here",
"OPENAI_DEFAULT_MODEL": "gpt-5-mini"
}
}
}
}Add to your MCP settings in Cursor:
Open Cursor Settings (Cmd/Ctrl + ,)
Search for "MCP" or go to Extensions → MCP
Add server configuration:
{
"mcpServers": {
"openai-websearch-mcp": {
"command": "uvx",
"args": ["openai-websearch-mcp"],
"env": {
"OPENAI_API_KEY": "your-api-key-here",
"OPENAI_DEFAULT_MODEL": "gpt-5-mini"
}
}
}
}Claude Code automatically detects MCP servers configured for Claude Desktop. Use the same configuration as above for Claude Desktop.
For local testing, use the absolute path to your virtual environment:
{
"mcpServers": {
"openai-websearch-mcp": {
"command": "/path/to/your/project/.venv/bin/python",
"args": ["-m", "openai_websearch_mcp"],
"env": {
"OPENAI_API_KEY": "your-api-key-here",
"OPENAI_DEFAULT_MODEL": "gpt-5-mini",
"PYTHONPATH": "/path/to/your/project/src"
}
}
}
}openai_web_searchIntelligent web search with reasoning model support.
Parameter | Type | Description | Default |
|
| The search query or question to search for | Required |
|
| AI model to use. Supports gpt-4o, gpt-4o-mini, gpt-5, gpt-5-mini, gpt-5-nano, o3, o4-mini |
|
|
| Reasoning effort level: low, medium, high, minimal | Smart default |
|
| Web search API version |
|
|
| Context amount: low, medium, high |
|
|
| Optional location for localized results |
|
Once configured, simply ask your AI assistant to search for information using natural language:
"Search for the latest developments in AI reasoning models using openai_web_search"
"Use openai_web_search with gpt-5 and high reasoning effort to provide a comprehensive analysis of quantum computing breakthroughs"
"Search for local tech meetups in San Francisco this week using openai_web_search"
The AI assistant will automatically use the openai_web_search tool with appropriate parameters based on your request.
Recommended: gpt-5-mini with reasoning_effort: "low"
Use Case: Fast iterations, real-time information, multiple quick queries
Benefits: Lower latency, cost-effective for frequent searches
Recommended: gpt-5 with reasoning_effort: "medium" or "high"
Use Case: Comprehensive analysis, complex topics, detailed investigation
Benefits: Multi-round reasoned results, no need for agent iterations
Model | Reasoning | Default Effort | Best For |
| ❌ | N/A | Standard search |
| ❌ | N/A | Basic queries |
| ✅ |
| Fast iterations |
| ✅ |
| Deep research |
| ✅ |
| Balanced approach |
| ✅ |
| Advanced reasoning |
| ✅ |
| Efficient reasoning |
# Install and run directly
uvx openai-websearch-mcp
# Or install globally
uvx install openai-websearch-mcp# Install from PyPI
pip install openai-websearch-mcp
# Run the server
python -m openai_websearch_mcp# Clone the repository
git clone https://github.com/yourusername/openai-websearch-mcp.git
cd openai-websearch-mcp
# Install dependencies
uv sync
# Run in development mode
uv run python -m openai_websearch_mcp# Clone and setup
git clone https://github.com/yourusername/openai-websearch-mcp.git
cd openai-websearch-mcp
# Create virtual environment and install dependencies
uv sync
# Run tests
uv run python -m pytest
# Install in development mode
uv pip install -e .Variable | Description | Default |
| Your OpenAI API key | Required |
| Default model to use |
|
# For uvx installations
npx @modelcontextprotocol/inspector uvx openai-websearch-mcp
# For pip installations
npx @modelcontextprotocol/inspector python -m openai_websearch_mcpIssue: "Unsupported parameter: 'reasoning.effort'" Solution: This occurs when using non-reasoning models (gpt-4o, gpt-4o-mini) with reasoning_effort parameter. The server automatically handles this by only applying reasoning parameters to compatible models.
Issue: "No module named 'openai_websearch_mcp'" Solution: Ensure you've installed the package correctly and your Python path includes the package location.
This project is licensed under the MIT License - see the LICENSE file for details.
🤖 Generated with Claude Code
🔥 Powered by OpenAI's Web Search API
🛠️ Built on the Model Context Protocol
Co-Authored-By: Claude noreply@anthropic.com
OpenAI Web Search with reasoning models.
For quick multi-round searches: Use 'gpt-5-mini' with reasoning_effort='low' for fast iterations.
For deep research: Use 'gpt-5' with reasoning_effort='medium' or 'high'. The result is already multi-round reasoned, so agents don't need continuous iterations.
Supports: gpt-4o (no reasoning), gpt-5/gpt-5-mini/gpt-5-nano, o3/o4-mini (with reasoning).
| Name | Required | Description | Default |
|---|---|---|---|
| input | Yes | The search query or question to search for | |
| model | No | AI model to use. Defaults to OPENAI_DEFAULT_MODEL env var or gpt-5-mini | |
| reasoning_effort | No | Reasoning effort level for supported models (gpt-5, o3, o4-mini). Default: low for gpt-5-mini, medium for others | |
| type | No | Web search API version to use | web_search_preview |
| search_context_size | No | Amount of context to include in search results | medium |
| user_location | No | Optional user location for localized search results |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden. It discloses that the result is 'already multi-round reasoned, so agents don't need continuous iterations', providing useful behavioral insight about the reasoning process. However, it doesn't mention potential rate limits, API requirements, or other operational characteristics. Since web search is inherently read-only, that aspect is implied.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is compact and front-loaded, with the main purpose stated in the first line. It then provides two clear usage scenarios and a list of supported models. Every sentence earns its place, with no redundant or filler content.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given that the tool has a rich schema and an output schema (which explains return values), the description doesn't need to cover those. It adequately covers the critical model selection guidance and reasoning effort recommendations, which are the non-obvious parts of using this tool. It's complete enough for an agent to decide when and how to invoke it.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema already includes descriptions for all 6 parameters (100% coverage), so the baseline is 3. The description adds extra meaning by explaining the relationship between models and reasoning_effort for different use cases, which goes beyond the schema's per-parameter descriptions. This helps the agent select appropriate model/effort combinations.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'OpenAI Web Search' which indicates the tool performs web searches, and mentions reasoning models. It doesn't use a specific verb like 'search' but the name and description together make the purpose clear. No sibling tools exist to differentiate from, so it doesn't need to distinguish itself.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description gives explicit guidance for different usage scenarios: 'For quick multi-round searches: Use gpt-5-mini with reasoning_effort='low'' and 'For deep research: Use gpt-5 with reasoning_effort='medium' or 'high''. This provides clear context on when to use specific model settings, though it doesn't mention when not to use the tool since there are no sibling tools.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v1.0.0openai_web_searchweb_searchweb_searchScored across 1 tool
With only one tool, there is no possibility of ambiguity or overlap between tools. The tool has a single, clear purpose of performing web searches with OpenAI models.
Since there is only one tool, naming consistency is inherently perfect. The tool name 'openai_web_search' follows a clear and descriptive pattern that would be consistent if more tools existed.
A single tool is generally too few for a server's purpose, as it limits functionality and flexibility. While the tool is well-described, the server's scope appears to be web search, which could benefit from additional tools for filtering, refining, or managing searches.
The server is severely incomplete for web search functionality. It lacks essential operations such as filtering results, handling pagination, saving or retrieving search history, or configuring search parameters beyond model selection. This will likely cause agent failures in complex search tasks.
One API key for 6 AI models. Pay-per-use. MCP protocol support with web search.
The best web search for your AI Agent
LLM-ready web search + instant answers + URL-to-clean-text fetch for agents and RAG.
Web search, browser automation, scraping, crawling and CAPTCHA solving for AI agents.