claude-agent-sdk

To build and troubleshoot Claude Agent SDK applications, install the SDK package, import it into your project, and validate the setup. Use the Python and TypeScript API references to configure sessions, manage permissions, and generate structured outputs.

Yes, you can integrate MCP servers and plugins with the Claude Agent SDK. The SDK provides examples and references for connecting MCP servers, plugins, Slash Commands, and Subagents to extend your agent-powered application with external tools and orchestration capabilities.

Subagents in the Claude Agent SDK are orchestrated components that handle specialized tasks within your application. They work by integrating with Skills, Plugins, and Slash Commands to enable complex agent workflows and multi-step task delegation.

Yes, the Claude Agent SDK supports structured outputs and custom system prompts. You can configure these features using the Python or TypeScript API to ensure agent responses adhere to specific data formats and behavioral guidelines for production-ready deployments.

The best way to manage sessions and permissions in the Claude Agent SDK is by using the core API configuration options for Python and TypeScript. The SDK guides you through setting up session states and permission controls to ensure secure, production-ready agent deployments.

To deploy production-ready coding agents using the Claude Agent SDK, follow the hosting best practices and integration patterns provided in the SDK. Configure sessions, permissions, and Subagents to build robust coding or business agents ready for production environments.