o3 Search

o3 Search is an MCP server that provides AI coding agents with direct access to OpenAI's high-end reasoning models and their web search capabilities. It connects AI clients to models such as o3, o4-mini, and gpt-5 via the official OpenAI API. Developers and software engineers use this server inside AI-assisted programming environments, such as Claude Code, to resolve niche engineering problems that require broad technical research. By exposing search and model consultation as standard tools, host agents can autonomously browse live internet resources, including GitHub issues and technical forums, to resolve errors, inspect updated API documentation, or assess architecture designs. Configurable parameters allow users to control context sizes, reasoning effort levels, API timeouts, and automatic retry attempts, providing fine-grained control over model behavior and query performance during automated software development workflows.

Category: Web Search & Research

Tags: information-retrieval, o3, openai, web search

Visit o3 Search

How to install and configure o3 Search

1. Verify that your OpenAI account is Tier 4 or organization-verified to ensure access to reasoning models. 2. Add the server using Claude Code: sh claude mcp add o3 -s user -e OPENAI_MODEL=o3 -e OPENAI_API_KEY=your-api-key -e SEARCH_CONTEXT_SIZE=medium -e REASONING_EFFORT=medium -- npx o3-search-mcp 3. Alternatively, register the server manually in your MCP client configuration JSON: json { "mcpServers": { "o3-search": { "command": "npx", "args": ["o3-search-mcp"], "env": { "OPENAI_API_KEY": "your-api-key", "OPENAI_MODEL": "o3", "SEARCH_CONTEXT_SIZE": "medium", "REASONING_EFFORT": "medium", "OPENAI_API_TIMEOUT": "300000", "OPENAI_MAX_RETRIES": "3" } } } } 4. Restart your MCP client to load the tool.

What you can do with o3 Search

• Debugging complex runtime errors by directing the agent to search GitHub issues and Stack Overflow discussions for obscure solutions. * Upgrading software libraries to newer versions by retrieving the latest unindexed documentation, breaking changes, and migration guides. * Replacing deprecated framework configuration options by identifying modern parameter equivalents and implementing the code changes automatically. * Conducting architectural design reviews by prompting reasoning models to evaluate collaborative editors or distributed system specifications.

Key facts

• https://github.com/yoshiko-pg/o3-search-mcp
• Web Search & Research
• information-retrieval, o3, openai, web search

Part of MCP Servers

How do I install o3 Search?

You can install o3 Search by running the claude mcp add command with npx o3-search-mcp, or by adding the server configuration to your MCP client JSON file. You must supply your OpenAI API key in the environment variables.

What can o3 Search do?

It allows MCP-compatible coding agents to search the web using OpenAI models like o3, o4-mini, and gpt-5. Agents can consult external sources to solve obscure bugs, review architectural designs, and inspect updated software documentation.

Which MCP clients work with o3 Search?

o3 Search works with MCP-compliant developer tools and interfaces that support Model Context Protocol servers over standard input and output, including Claude Code and desktop client environments.

What OpenAI API tier is required for o3 Search?

To use the o3 model via the OpenAI API, your account must be upgraded to Tier 4 or verified as an organization. Using an API key without o3 permissions will cause server requests to return errors.

Can I configure reasoning effort and context size in o3 Search?

Yes. You can customize the server behavior using environment variables such as SEARCH_CONTEXT_SIZE and REASONING_EFFORT, which accept low, medium, or high values. You can also configure request timeouts and retry counts.