The Claude Cookbook, an official repository hosted at platform.claude.com, serves as an extensive collection of guides, tutorials, and code patterns for developers building applications with Claude. The library covers a wide range of agentic workflows and integration techniques, with documentation dating back to March 2024. Recent additions as of June 2026 focus on advanced agentic capabilities, such as reproducing benchmark scores for DeepSearchQA and BrowseComp, implementing classifier fallback patterns for safety, and orchestrating multi-agent systems.

The repository organizes content into specific categories including agent patterns, tool use, retrieval-augmented generation (RAG), and multimodal processing. Developers can access technical resources for programmatic tool calling (PTC), which optimizes token consumption and latency, alongside methods for dynamic tool discovery through semantic embeddings. Other documentation outlines strategies for automatic context compaction, image analysis via cropping tools, and frontend design optimization.

Specific guides detail the construction of specialized agents for tasks like cybersecurity vulnerability detection, site reliability engineering (SRE) incident response, and data analysis. The cookbook also provides instructions for Claude Managed Agents, covering features like session memory, prompt versioning, and vault-backed credential management using the Model Context Protocol (MCP). The resource includes migration paths for developers moving from other agent SDKs and best practices for cost-effective administration of API usage.

Content is maintained through contributions from Anthropic and various engineers, offering modular patterns for teams to build, deploy, and evaluate their own AI workflows. These patterns range from simple classification tasks and summarization techniques to complex, multi-agent orchestrations that allow specialist teams to perform research, pricing, and automated report generation.

The Claude Cookbook, an official repository hosted at platform.claude.com, serves as an extensive collection of guides, tutorials, and code patterns for developers building applications with Claude. The library covers a wide range of agentic workflows and integration techniques, with documentation dating back to March 2024. Recent additions as of June 2026 focus on advanced agentic capabilities, such as reproducing benchmark scores for DeepSearchQA and BrowseComp, implementing classifier fallback patterns for safety, and orchestrating multi-agent systems.

The repository organizes content into specific categories including agent patterns, tool use, retrieval-augmented generation (RAG), and multimodal processing. Developers can access technical resources for programmatic tool calling (PTC), which optimizes token consumption and latency, alongside methods for dynamic tool discovery through semantic embeddings. Other documentation outlines strategies for automatic context compaction, image analysis via cropping tools, and frontend design optimization.

Specific guides detail the construction of specialized agents for tasks like cybersecurity vulnerability detection, site reliability engineering (SRE) incident response, and data analysis. The cookbook also provides instructions for Claude Managed Agents, covering features like session memory, prompt versioning, and vault-backed credential management using the Model Context Protocol (MCP). The resource includes migration paths for developers moving from other agent SDKs and best practices for cost-effective administration of API usage.

Content is maintained through contributions from Anthropic and various engineers, offering modular patterns for teams to build, deploy, and evaluate their own AI workflows. These patterns range from simple classification tasks and summarization techniques to complex, multi-agent orchestrations that allow specialist teams to perform research, pricing, and automated report generation.