Direct answer
What is Claude Agent SDK?
The Claude Agent SDK is a developer library in Python and TypeScript that provides Claude Code's autonomous agent loop, built-in tools, and context management directly inside your own application process.
The Claude Agent SDK packages the foundational mechanics of Claude Code into an embeddable library for Python and TypeScript. Rather than requiring developers to manually build, prompt, and resolve iterative tool-calling sequences, the SDK runs the planning and execution loop natively within your own process. The framework provides built-in tools to read, write, and edit files, execute shell commands, and conduct web searches. It also includes support for the Model Context Protocol (MCP) to access external APIs and data sources, along with features such as subagents, lifecycle hooks, permission controls, session persistence, and file-change checkpointing. Unlike managed cloud agent services, the Agent SDK executes in your own infrastructure, giving teams direct authority over security, permissions, sandboxing, and session storage while authenticating through standard Anthropic API keys.
Unlike raw API clients that require you to build the tool-calling loop from scratch, the Agent SDK embeds Claude Code's complete autonomous agent loop, filesystem tools, subagents, and MCP support directly into Python and TypeScript codebases.
Product capabilities
Key features
Autonomous Agent Loop
Runs task planning, tool selection, and iterative execution natively inside your own application process.
Built-in System & Code Tools
Comes equipped with tools to read, write, and edit project files, execute terminal commands, and perform web searches.
Model Context Protocol (MCP)
Connects external data sources, enterprise tools, and custom services via the standardized Model Context Protocol.
Subagent Delegation
Spawns focused subagents to isolate tasks and execute dynamic, multi-agent workflows.
Lifecycle Hooks & Approvals
Intercepts agent actions at key stages and configures granular permissions to determine which actions run automatically versus requiring manual approval.
Sessions & Checkpointing
Maintains conversation context across exchanges, allows sessions to be resumed or forked, and rewinds file modifications when needed.
Observability & Metrics
Supports OpenTelemetry integration along with token cost, usage, and internal task tracking.
Practical fit
Who should use Claude Agent SDK?
Automated Bug Resolution
Building programmatic background agents that inspect error logs, locate relevant repository files, and apply automated fixes.
In-House Developer Assistants
Embedding Claude's coding and system execution capabilities directly into proprietary CLI utilities or internal developer platforms.
Complex Codebase Migrations
Orchestrating specialized subagents to coordinate dependency upgrades and multi-file code modifications under explicit permission controls.
Editorial assessment
Pros and limitations
Where it is strong
- Embeds the proven Claude Code execution loop and filesystem tools into your own code.
- First-class support for MCP servers, subagents, and custom lifecycle hooks.
- Full control over local execution, sandboxing, and session persistence.
Where to be careful
- Requires managing your own process execution environment and sandboxing.
- Available natively only for Python and TypeScript; other languages must wrap the CLI subprocess.
- Third-party products cannot use consumer claude.ai logins or rate limits and must supply API keys.
Commercial context
Claude Agent SDK pricing
At the review date of 2026-09-08, the Claude Agent SDK is governed by Anthropic's Commercial Terms of Service and uses API key authentication. Token pricing and infrastructure costs depend on standard Anthropic API usage rates; verify current rates on Anthropic's official pricing page.
Pricing, limits, taxes, model access, and regional availability can change. Verify the purchase-critical details on the official pricing page linked under Sources.
Transparent ranking
Why Claude Agent SDK scores 76.5
Each factor is scored on a 100-point scale, then combined using the public ToolsRank weights. Engagement and momentum stay at a neutral baseline until measured signals exist, so no tool can gain or lose position from numbers nobody recorded.
Compatibility
Languages, platforms, and integrations
Languages
- Python
- TypeScript
Integrations & surfaces
- Anthropic API
- Model Context Protocol (MCP)
- OpenTelemetry
Community
Reviews and questions
No approved member reviews yet. Editorial factors above are the only rating on this page.
Reviews and questions come from Google-signed members and are checked by an editor before they appear.
Frequently asked
Claude Agent SDK FAQ
How does the Agent SDK differ from the standard Anthropic Client SDK?+
The Client SDK offers basic API connectivity, requiring you to construct and manage the loop that handles tool calls and model responses yourself. The Agent SDK provides the complete agent loop, file editing tools, and context management engine that powers Claude Code.
What programming languages are supported by the Agent SDK?+
The Agent SDK is available as a native library for Python and TypeScript. To drive the agent loop from other programming environments, developers can run the Claude Code CLI as a subprocess using the -p flag and JSON output format.
Can I allow end users to log in with their claude.ai accounts?+
No. Unless previously approved by Anthropic, third-party developers are not permitted to offer claude.ai consumer logins or consumer rate limits. Applications built with the SDK must authenticate via Anthropic API keys.
Where does the Agent SDK execute terminal commands and tool calls?+
The Agent SDK runs directly within your own application process and infrastructure. You are responsible for hosting, securing the environment, and applying sandboxing to govern file access and terminal command execution.
