o3-search MCP
The o3-search MCP server provides AI-powered web search capabilities using OpenAI's o3 model.
Natural Language Queries: Accepts text questions or search requests in English
Advanced Search: Finds the latest information and helps troubleshoot errors
Customizable Context: Allows adjusting search context size (
low,medium,high)Reasoning Effort Control: Enables setting reasoning effort level (
low,medium,high)API Integration: Compatible with OpenAI's API, with configurable timeout and retry settings
Complex Problem Solving: Assists in consulting about complex problems
Provides web search capabilities using OpenAI's o3 model, enabling AI agents to perform text-based web searches with configurable context size and reasoning effort
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., "@o3-search MCP search for the latest React 19 breaking changes and migration guide"
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.
o3-search-mcp (gpt-5, o4-mini support)
MCP server that enables the use of OpenAI's high-end models and their powerful web search capabilities. By registering it with any AI coding agent, the agent can autonomously consult with OpenAI models to solve complex problems.
Use Cases
(Although called o3 to match the MCP name, you can specify gpt-5 or o4-mini via env for the model to use)
🐛 When you're stuck debugging
o3's web search can scan a wide range of sources, including GitHub issues and Stack Overflow, significantly increasing the chances of resolving niche problems. Example prompts:
> I'm getting the following error on startup, please fix it. If it's too difficult, ask o3.
> [Paste error message here]> The WebSocket connection isn't working. Please debug it. If you don't know how, ask o3.📚 When you want to reference the latest library information
You can get answers from the powerful web search even when there's no well-organized documentation. Example prompts:
> I want to upgrade this library to v2. Proceed while consulting with o3.> I was told this option for this library doesn't exist. It might have been removed. Ask o3 what to specify instead and replace it.🧩 When tackling complex tasks
In addition to search, you can also use it as a sounding board for design. Example prompts:
> I want to create a collaborative editor, so please design it. Also, ask o3 for a design review and discuss if necessary.Also, since it's provided as an MCP server, the AI agent may decide on its own to talk to o3 when it deems it necessary, without any instructions from you. This will dramatically expand the range of problems it can solve on its own!
Related MCP server: 4o-mini-search-mcp
Installation
npx (Recommended)
Claude Code:
$ claude mcp add o3 \
-s user \ # If you omit this line, it will be installed in the project scope
-e OPENAI_MODEL=o3 \ # o4-mini, gpt-5 also available
-e OPENAI_API_KEY=your-api-key \
-e SEARCH_CONTEXT_SIZE=medium \
-e REASONING_EFFORT=medium \
-e OPENAI_API_TIMEOUT=300000 \
-e OPENAI_MAX_RETRIES=3 \
-- npx o3-search-mcpjson:
{
"mcpServers": {
"o3-search": {
"command": "npx",
"args": ["o3-search-mcp"],
"env": {
"OPENAI_API_KEY": "your-api-key",
// Optional: o3, o4-mini, gpt-5 (default: o3)
"OPENAI_MODEL": "o3",
// Optional: low, medium, high (default: medium)
"SEARCH_CONTEXT_SIZE": "medium",
"REASONING_EFFORT": "medium",
// Optional: API timeout in milliseconds (default: 300000)
"OPENAI_API_TIMEOUT": "300000",
// Optional: Maximum number of retries (default: 3)
"OPENAI_MAX_RETRIES": "3"
}
}
}
}Local Setup
If you want to download the code and run it locally:
git clone git@github.com:yoshiko-pg/o3-search-mcp.git
cd o3-search-mcp
pnpm install
pnpm buildClaude Code:
$ claude mcp add o3 \
-s user \ # If you omit this line, it will be installed in the project scope
-e OPENAI_MODEL=o3 \ # o4-mini, gpt-5 also available
-e OPENAI_API_KEY=your-api-key \
-e OPENAI_MODEL=o3 \
-e SEARCH_CONTEXT_SIZE=medium \
-e REASONING_EFFORT=medium \
-e OPENAI_API_TIMEOUT=300000 \
-e OPENAI_MAX_RETRIES=3 \
-- node /path/to/o3-search-mcp/build/index.jsjson:
{
"mcpServers": {
"o3-search": {
"command": "node",
"args": ["/path/to/o3-search-mcp/build/index.js"],
"env": {
"OPENAI_API_KEY": "your-api-key",
// Optional: o3, o4-mini, gpt-5 (default: o3)
"OPENAI_MODEL": "o3",
// Optional: low, medium, high (default: medium)
"SEARCH_CONTEXT_SIZE": "medium",
"REASONING_EFFORT": "medium",
// Optional: API timeout in milliseconds (default: 300000)
"OPENAI_API_TIMEOUT": "300000",
// Optional: Maximum number of retries (default: 3)
"OPENAI_MAX_RETRIES": "3"
}
}
}
}Environment Variables
Environment Variable | Options | Default | Description |
| Required | - | OpenAI API Key |
| Optional |
| Model to useValues: |
| Optional |
| Controls the search context sizeValues: |
| Optional |
| Controls the reasoning effort levelValues: |
| Optional |
| API request timeout in millisecondsExample: |
| Optional |
| Maximum number of retries for failed requestsThe SDK automatically retries on rate limits (429), server errors (5xx), and connection errors |
Notes
To use the o3 model from the OpenAI API, you need to either raise your tier to 4 or verify your organization. If you register an API key that is not yet enabled for o3 with this MCP, calls will result in an error. Reference: https://help.openai.com/en/articles/10362446-api-access-to-o1-o3-and-o4-models
Available Tools
1 toolo3-searchA
An AI agent with advanced web search capabilities. Useful for finding the latest information, troubleshooting errors, and discussing ideas or design challenges. Supports natural language queries.
Inputs: input (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 1 tool
Only one tool exists, so there is no possibility of confusion between tools.
With a single tool, naming is inherently consistent; the name 'o3-search' clearly indicates its search function.
Having only one tool feels thin for a search server, but the tool is non-trivial and covers general web search, so it's borderline acceptable.
The single tool covers web search comprehensively with natural language support, though more specialized search options could be beneficial.
Maintenance
Related MCP Connectors
Web search API and MCP server for developers building AI agents.
Driflyte MCP server which lets AI assistants query topic-specific knowledge from web and GitHub.
Free web search for AI agents. No API key required. Hosted MCP in active development.
Related MCP Servers
- FlicenseCqualityDmaintenanceAn MCP protocol server that enables web search functionality using the Tavily API, allowing AI assistants to perform internet searches in real-time.44-
- FlicenseNot gradedqualityNot gradedmaintenanceAn MCP server that enables AI models to search the web using OpenAI's 4o-mini Search model, allowing access to up-to-date information for just a few cents per search.3 npm2-
- AlicenseAqualityDmaintenanceAn MCP server that provides AI assistants with web search and intelligence capabilities via the ihyee API. It allows users to search the web, fetch extracted content from URLs, and perform full browser rendering for JavaScript-heavy websites.3MIT
- AlicenseAqualityFmaintenanceAn MCP server that provides real-time web search to AI agents via a pay-per-search USDC microtransaction system.579 npm1MIT