Claude Agent SDK
Build Python and TypeScript agents with Claude Code’s tools, sessions, and execution loop.
- Tool type
- Agent development SDK
- Made for
- Agent and application developers
- Runs on
- Python or TypeScript
- Runs in infrastructure you operate
What it does
The Claude Agent SDK lets developers embed Claude Code’s agent loop in their own Python or TypeScript application. Instead of writing every step of a tool-calling loop, you give the agent a goal and control the tools and resources it may use. The SDK supplies file operations, command execution, web search, context management, and sessions that can continue across exchanges.
It is useful when your application needs an agent to do work: investigate a repository, assemble research, process files, or coordinate specialized subtasks. Hooks let your code respond at key moments, permissions determine which actions need approval, and MCP connects other tools and data. You operate the process and remain responsible for its runtime, credentials, and access boundaries.
This is a developer library, so getting started means installing a package and writing an application. The quickstart requires Node.js 18.14 or later, or Python 3.10 or later, plus an Anthropic account and API key. Model and tool usage are billed separately; a personal Claude subscription is not a general entitlement to authenticate customers of your own SDK-based product.
What it can do
- Embed an agent loop with built-in tools
- Create and resume agent sessions
- Delegate tasks to specialized subagents
- Add hooks and approval rules
- Connect external tools through MCP
Works with
- Python
- TypeScript
- MCP servers
- Claude Code runtime
Example workflows
- A repository maintenance agent
- Wrap the SDK in an internal service that reads a repo, proposes a fix, runs tests, and returns a diff for review.
- A controlled document workflow
- Allow only approved folders and tools, then let an agent gather files and produce a structured deliverable.
Before you choose
- You host the process; this is different from Anthropic’s hosted Managed Agents.
- Third-party products should use the documented API authentication rather than customers’ personal Claude login.
Listed under
Get started
Node.js 18.14+ or Python 3.10+
- Install the Python or TypeScript SDK package in a project.
- Create an API key and configure it securely in the application environment.
- Choose allowed tools and permissions, then run a small task before expanding access.
Plans and prices
Prices in USDExample standard model rate, not a flat SDK subscription or total task cost. Other models, tools, and your hosting can cost more.
Source: platform.claude.com/docs/en/about-claude/pricing, checked 28 Oct 2026. Prices can change; check before you buy.
Sources for this overview
Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.
- Agent SDK overview - Claude Code Docs ↗code.claude.com
- Quickstart - Claude Code Docs ↗code.claude.com
- Pricing - Claude Platform Docs ↗platform.claude.com
Ratings and reviews
No ratings for Claude Agent SDK yet.
What’s changed
Free accountSignals
Free accountData and security
Free accountCompare with similar tools
Free accountWhere this comes from
How we research- Company website
- code.claude.com: product, pricing and policy pages
- YC directory mirror
- Batch and company name
- Web search
- Confirms the official site
- Social accounts
- Accounts the company links, and its posts on X
- Human review
- Checked by a person
- knowledge_base
Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.