Building with extended thinking - Claude Docs

Example: Passing thinking blocks with tool results Tool use without interleaved thinking Tool use with interleaved thinking System prompt caching (preserved when thinking changes) Messages caching (invalidated when thinking changes) Example: Working with redacted thinking blocks

Configure manual extended thinking with a fixed budget_tokens budget on Claude models that support it, and migrate to adaptive thinking. Extended thinking in manual mode gives you direct control over how much Claude thinks. You set a thinking token budget on each request with thinking: {type: "enabled", budget_tokens: N}, and Claude thinks against that budget before it starts its final answer. Manual mode remains useful when your workload requires predictable latency or precise control over thinking costs. This page covers how to set and tune the budget, how manual mode interacts with…

related reading

• Claude's extended thinking \ Anthropicanthropic.com
• Prompting best practicesdocs.anthropic.com
• The "think" tool: Enabling Claude to stop and think \ Anthropicanthropic.com
• Maximizing the value of your Claude Code sessionsclaude.com
• Tracing the thoughts of a large language model \ Anthropicanthropic.com
• A global workspace in language models \ Anthropicanthropic.com
• Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Labthinkingmachines.ai
• A Guide to Claude Code 2.0 and getting better at using coding agents – sankalp's blogsankalp.bearblog.dev
• Best practices for Claude Code - Claude Code Docsanthropic.com
• Inkling: Our Open-Weights Model - Thinking Machines Labthinkingmachines.ai
• ClaudeLog - Claude Code Docs, Guides, Tutorials & Best Practicesclaudelog.com
• Claude Code Cheat Sheetcc.storyfox.cz