Claude Agent SDK Overview
by Anthropic
Build production AI agents on the same tools, agent loop, and context management that power Claude Code.
Overview
The Claude Agent SDK Overview is Anthropic's entry-point documentation for the Claude Agent SDK (published as claude-agent-sdk on PyPI and @anthropic-ai/claude-agent-sdk on npm), which exposes the same tools, agent loop, and context management that power Claude Code as a library in Python and TypeScript. Out of the box, agents get built-in tools (Read, Write, Edit, Bash, Glob, Grep, WebSearch, WebFetch, Monitor, AskUserQuestion), lifecycle hooks (PreToolUse, PostToolUse, SessionStart, and more), subagents for delegating focused subtasks, first-class Model Context Protocol (MCP) client support, fine-grained tool permissions, and resumable/forkable sessions. It also loads Claude Code's filesystem configuration — Skills, slash commands, CLAUDE.md memory, and plugins — and documents authentication via the Anthropic API, Amazon Bedrock, Google Vertex, and Microsoft Foundry, plus how the SDK differs from the Client SDK, the Claude Code CLI, and hosted Managed Agents.
At a Glance
- Topic
- Agentic
- Level
- Intermediate
- Format
- Documentation
- Cost
- Free
- Duration
- Self-paced reference docs (Python & TypeScript)
- Provider
- Anthropic
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Run an autonomous agent loop with query() in Python or TypeScript without implementing your own tool-execution loop
- ✓Use built-in tools (Read, Write, Edit, Bash, Glob, Grep, WebSearch, WebFetch) so agents can act on a filesystem and the web immediately
- ✓Add lifecycle hooks (PreToolUse/PostToolUse/Stop/SessionStart) to validate, log, block, or transform agent behavior
- ✓Delegate work to subagents with their own context and track them via parent_tool_use_id
- ✓Connect external systems (databases, browsers, APIs) through the Model Context Protocol and load Skills, slash commands, and CLAUDE.md config
- ✓Control tool permissions and maintain, resume, and fork sessions for multi-step work
Highlights
- •First-party Anthropic docs kept current with the SDK, with runnable Python and TypeScript examples side by side
- •Same harness as Claude Code, so local prototypes translate directly to production automation and CI/CD
- •Covers subagents, hooks, MCP, permissions, and sessions in one place, plus comparisons to the Client SDK and Managed Agents
Who It’s For
Best For
- ✓AI engineers building autonomous coding, research, or ops agents in Python or TypeScript
- ✓Teams moving an agent from a Claude Code prototype to production or CI/CD
- ✓Developers who want built-in tool execution and MCP support without wiring their own tool loop
Prerequisites
- •Working Python 3.10+ or Node.js/TypeScript proficiency
- •An Anthropic API key (or Bedrock/Vertex/Foundry credentials)
- •Familiarity with LLM tool use / agent concepts
FAQ
What is Claude Agent SDK Overview?
Official Anthropic documentation for the Claude Agent SDK, the Python and TypeScript library for building production AI agents that autonomously read files, run commands, search the web, and edit code. For engineers who want Claude Code's harness as a programmable library rather than a CLI.
Is Claude Agent SDK Overview free?
Claude Agent SDK Overview is free to access.
What level is Claude Agent SDK Overview for?
Claude Agent SDK Overview is aimed at a intermediate audience. Recommended background: Working Python 3.10+ or Node.js/TypeScript proficiency, An Anthropic API key (or Bedrock/Vertex/Foundry credentials), Familiarity with LLM tool use / agent concepts.
How long does Claude Agent SDK Overview take?
Expect roughly Self-paced reference docs (Python & TypeScript). Most learners work through it at their own pace.
What will I learn from Claude Agent SDK Overview?
You'll learn: Run an autonomous agent loop with query() in Python or TypeScript without implementing your own tool-execution loop; Use built-in tools (Read, Write, Edit, Bash, Glob, Grep, WebSearch, WebFetch) so agents can act on a filesystem and the web immediately; Add lifecycle hooks (PreToolUse/PostToolUse/Stop/SessionStart) to validate, log, block, or transform agent behavior; Delegate work to subagents with their own context and track them via parent_tool_use_id; Connect external systems (databases, browsers, APIs) through the Model Context Protocol and load Skills, slash commands, and CLAUDE.md config; Control tool permissions and maintain, resume, and fork sessions for multi-step work.