Step-by-step tutorials on building voice and chat AI agents using OpenAI Agents SDK, Realtime API, function calling, multi-agent orchestration, and production deployment patterns.
9 of 1313 articles
Understand MCP, the open protocol for connecting AI agents to external tools and data sources, including its architecture, five transport types, and how to build your first MCP-connected agent.
Connect agents to remote MCP tool servers using MCPServerStreamableHTTP with authentication headers, timeout configuration, retry policies, tool caching, and production deployment patterns.
Use HostedMCPTool to run MCP tools server-side on OpenAI's infrastructure with zero callback overhead, including tool_config structure, approval options, and GitMCP for repository access.
Dramatically reduce agent startup latency by caching MCP tool definitions with cache_tools_list, implementing cache invalidation strategies, and benchmarking the performance gains in production agents.
Use MCP prompt resources to dynamically load and parameterize agent instructions from external servers, enabling centralized prompt management with list_prompts and get_prompt.
Use MCPServerManager to orchestrate multiple MCP server connections with automatic failure detection, reconnection strategies, and health monitoring using active_servers, failed_servers, and drop_failed_servers.
Learn how the OpenAI Agents SDK automatically traces every agent run with agent_span, generation_span, and function_span, and how to visualize traces in the OpenAI dashboard for debugging.
Learn how to use the trace() context manager, custom_span(), and manual span lifecycle to build detailed, hierarchical trace visualizations for complex multi-step agent workflows.
