Also known as: Claude Agent SDK, Anthropic Claude Agent SDK, Claude Code Hooks SDK
The programmatic TypeScript surface for building agents on top of Claude. Covers the 12 hook events (PreToolUse, PostToolUse, UserPromptSubmit, Stop, SubagentStop, PreCompact, SessionStart, SessionEnd, Notification, PermissionRequest, SubagentStart, PostToolUseFailure), the HookCallback signature, hookSpecificOutput decisions (allow/deny/ask + updatedInput), and how programmatic hooks compare to declarative Claude Code settings.json hooks.
Canonical resource
/blogs/anthropic-agent-sdk-hooks-typescript-referenceAnthropic Agent SDK Hooks: TypeScript Reference
Programmatic Anthropic Agent SDK hooks reference. 12 events, full TypeScript signatures, PreToolUse, PostToolUse, Stop, SubagentStop, UserPromptSubmit with worked examples.
May 23, 2026
Claude Code Hooks: 6 Production Patterns (2026)
Six production Claude Code hooks patterns: auto-format, file protection, security gates, type-checks, prompt review. Real settings.json examples.
Feb 14, 2026
Context Engineering Beyond CLAUDE.md: The 5-Layer Hierarchy
CLAUDE.md is just layer one. The five-layer context hierarchy, memory patterns, and subagent strategies that separate productive AI coding from prompt guessing. With working examples.
Feb 14, 2026
Want to implement anthropic agent sdk in your product?
Talk to our teamOur most-cited deep dives on AI search visibility, plus what we shipped this month.
Learn the difference between SEO, AEO, and GEO, and how brands can move from ranking in search to being cited in answers and recommended by AI engines.
Most AI visibility scores are opaque grades you cannot defend. Here is how Radar scores AI visibility in three separated layers, with a trail behind every number.
Google Preferred Sources is the one AI visibility lever your audience controls, not the model. Here is how it works, who qualifies, and how to earn it.