Skills & Tools

This cookbook demonstrates Claude 3.7 Sonnet's extended thinking feature, which provides transparent step-by-step reasoning with budget management. Extended thinking enables Claude to show its internal reasoning process through thinking content blocks before delivering final answers. The guide covers setup, basic examples, streaming with thinking, token counting, redacted thinking, and error handling with practical Python code examples.

Claude 3.7 Sonnet may not make parallel tool calls by default even when enabled. This cookbook demonstrates a workaround using a "batch tool" meta-pattern that wraps multiple tool invocations, encouraging the model to call multiple tools simultaneously in a single response. By introducing a batch_tool that accepts an array of tool invocations, developers can improve latency and efficiency by processing multiple tool calls in parallel rather than sequentially.

Speculative prompt caching is a technique that reduces time-to-first-token (TTFT) by warming up Claude's cache while users are still formulating their queries. Instead of waiting until a user submits their question to load context into the cache, the system begins cache warming immediately when the user starts typing. This cookbook demonstrates the pattern using SQLite source code as example context, showing how to implement it with the Anthropic Python SDK using async operations and prompt caching controls.

This cookbook demonstrates how to build AI agents with persistent memory using Claude's memory tool and context editing capabilities. It addresses challenges of long-running agents losing learned patterns between sessions and context window overflow by implementing cross-conversation learning and automatic context management. The guide covers practical implementations for use cases like code review assistants, research assistants, and customer support bots, with setup instructions and best practices for memory security and organization.

This cookbook article introduces the Claude Agent SDK for building multi-agent systems, using a Chief of Staff agent for a startup as the primary example. It demonstrates key features including persistent memory via CLAUDE.md files, bash tool execution for Python scripts, and coordination of specialized subagents. The article progressively builds a comprehensive agent system that aggregates insights from multiple domains to provide executive summaries and actionable recommendations for a CEO managing a $10M Series A startup.

This cookbook teaches how to build a research agent using Claude Agent SDK with WebSearch tool for autonomous information gathering and synthesis. The guide demonstrates creating a functional research agent in just a few lines of code, then progresses to production improvements including conversation memory, system prompts for specialized behavior, and multimodal research capabilities. The agent autonomously decides when and how to search, follows promising leads, and synthesizes findings without predefined workflows.

This cookbook guide teaches developers how to create, deploy, and manage custom skills for Claude that extend its capabilities with organization-specific workflows and domain knowledge. Custom skills are specialized expertise packages bundled as markdown files, scripts, and resources that codify organizational knowledge, ensure consistency, and automate complex workflows while maintaining privacy. The guide covers skill architecture, SKILL.md requirements, progressive disclosure for token optimization, and provides utility functions for skill management including creation, listing, and deletion.

This cookbook teaches how to build financial dashboards, portfolio analytics, and automated reporting workflows using Claude's Excel, PowerPoint, and PDF skills. It covers three main use cases: creating comprehensive financial models with formulas and charts in Excel, generating executive presentations from financial data, and automating multi-format reporting pipelines. The guide includes setup instructions, data loading examples, and helper functions for working with the Anthropic SDK to create professional financial documents directly through Claude's interface.

This cookbook demonstrates building a low-latency voice assistant by combining ElevenLabs (speech-to-text and text-to-speech) with Claude for intelligent responses. The guide covers installation, API setup, and crucially, latency optimization techniques including Claude's streaming API and sentence-by-sentence TTS synthesis. Performance measurements show streaming reduces perceived latency by ~31% compared to non-streaming approaches, with TTS first-chunk delivery in 0.39 seconds.

This cookbook demonstrates building an autonomous SRE incident response agent using the Claude Agent SDK with read-write MCP tools for safe infrastructure access. The agent can investigate incidents by querying metrics and logs, diagnose root causes, apply remediations by editing configs and restarting services, and document post-mortems. The pattern uses a subprocess-based MCP server with scoped tool access, clear tool descriptions, and human-in-the-loop workflows to enable autonomous yet controlled incident response.

This guide introduces Claude Skills, specialized capability packages that enable document creation, data analysis, and workflow automation through Excel, PowerPoint, and PDF generation. Skills use a three-tier progressive loading model (metadata → full instructions → linked files) to optimize token usage and efficiency. The tutorial covers environment setup, API configuration, skill discovery, and practical quick-start examples for Excel, PowerPoint, and PDF workflows.

This cookbook demonstrates how to build a crop tool for Claude to analyze images with greater detail by zooming into specific regions. The tool uses normalized coordinates (0-1) to let Claude request cropped sections of charts, documents, and diagrams without needing to know pixel dimensions. An agentic loop handles tool calls iteratively until Claude provides a final answer, enabling precise analysis of small text, chart comparisons, and technical details.

This cookbook demonstrates how to scale Claude applications from dozens to thousands of tools using semantic embeddings for dynamic tool discovery. Instead of front-loading all tool definitions (which consumes context and increases latency), the approach provides Claude with a single tool_search tool that returns relevant capabilities on demand, reducing context usage by 90%+. The guide walks through implementing client-side semantic search using SentenceTransformer embeddings to match user queries with appropriate tools from large libraries, making it practical for production applications managing extensive tool ecosystems.

Programmatic Tool Calling (PTC) enables Claude to write and execute code that calls tools directly within the code execution environment, eliminating round-trips through the model for each tool invocation. This approach significantly reduces latency and token consumption, especially when dealing with large datasets or sequential tool dependencies. The cookbook demonstrates PTC using a team expense management API scenario where Claude analyzes employee expenses across multiple tool calls, filters irrelevant data programmatically, and identifies budget overages without excessive context window usage.

This content discusses strategies for optimizing Gemini 3 performance for specific use cases, achieving 10x improvements. It references prompt engineering best practices and introduces the Superdesign agent as a tool for implementation. The video likely covers techniques for tailoring large language models to particular applications through effective prompting and agent-based workflows.

LangChain v0.3.28 is a maintenance and security release that addresses a critical ReDoS vulnerability (CVE-2024-58340) in MRKL and ReAct action regex patterns. The release includes improvements to UUID7 for run IDs, enhanced OpenAI streaming support, better Anthropic model integration, and extensive code quality improvements including Pydantic deprecation fixes, Ruff linting enhancements, and documentation standardization.

Microsoft AutoGen python-v0.6.2 introduces streaming tools with `BaseStreamTool` and `StreamWorkbench` support, adds `tool_choice` parameter to `ChatCompletionClient`, and enables `AssistantAgent` with an inner tool calling loop via `max_tool_iterations`. Additional features include OpenTelemetry GenAI traces, Mem0 memory extension, `output_task_messages` flag, GraphFlow improvements, and support for latest models like Gemini 2.5 Flash.

Microsoft AutoGen python-v0.6.4 release includes improvements to GraphFlow for better state management and task resumption, enhancements to Workbench implementations with tool name/description overrides, and fixes for function calling support and conversation context pollution. The release also adds new features like Qwen2.5VL support, DuckDuckGo search tools, and improved documentation generation.

CrewAI 1.10.0 introduces enhanced MCP tool resolution, improved JSON argument parsing, user input handling in Flows, and migrates the CLI HTTP client from requests to httpx. The release includes 20+ bug fixes addressing tool validation, schema handling, callback execution, and race conditions, plus documentation updates for security guidance and 96 new integration actions.

langchain-text-splitters version 1.1.1 was released with bug fixes including prevention of JSFrameworkTextSplitter mutating internal separators, resolution of a Server-Side Request Forgery vulnerability in HTMLHeaderTextSplitter, and fixes to HTMLSemanticPreservingSplitter. The release also includes dependency updates, code style improvements with ruff linting, and new features like model_kwargs support for SentenceTransformersTokenTextSplitter.