# Claude Agent Sdk

> |

- **Type:** Skill
- **Install:** `agentstack add skill-xiaolai-anthropic-docs-claude-agent-sdk`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [xiaolai](https://agentstack.voostack.com/s/xiaolai)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [xiaolai](https://github.com/xiaolai)
- **Source:** https://github.com/xiaolai/anthropic-docs/tree/main/skills/claude-agent-sdk
- **Website:** https://github.com/xiaolai/autoupdated-anthropic-documentation-knowledge

## Install

```sh
agentstack add skill-xiaolai-anthropic-docs-claude-agent-sdk
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Claude Agent SDK Reference

| | TypeScript | Python |
|---|---|---|
| **Version** | v0.3.159 | v0.2.87 |
| **Package** | `@anthropic-ai/claude-agent-sdk` | `claude-agent-sdk` (PyPI) |
| **Docs** | [TypeScript SDK](https://platform.claude.com/docs/en/agent-sdk/typescript) | [Python SDK](https://platform.claude.com/docs/en/agent-sdk/python) |
| **Repo** | [claude-agent-sdk-typescript](https://github.com/anthropics/claude-agent-sdk-typescript) | [claude-agent-sdk-python](https://github.com/anthropics/claude-agent-sdk-python) |
| **Full reference** | [SKILL-typescript.md](SKILL-typescript.md) | [SKILL-python.md](SKILL-python.md) |

## When you detect the user's language

- Working with `.ts` files, TypeScript imports, or `npm`/`node` → read `SKILL-typescript.md`
- Working with `.py` files, Python imports, or `pip`/`python` → read `SKILL-python.md`
- Ambiguous or multi-language → read both surface files

## Cross-Language Naming Map

| Concept | TypeScript | Python |
|---------|-----------|--------|
| One-shot query | `query(options)` | `query(options)` |
| Stateful client | N/A (query manages state) | `ClaudeSDKClient` |
| Options type | `Options` interface | `ClaudeAgentOptions` dataclass |
| Tool definition | `tool(name, schema, handler)` | `@tool(name, desc, schema)` decorator |
| MCP server factory | `createSdkMcpServer()` | `create_sdk_mcp_server()` |
| Permission callback | `canUseTool` | `can_use_tool` |
| Permission mode | `permissionMode: "..."` | `permission_mode="..."` |
| Hook registration | `hooks: { PreToolUse: [...] }` | `hooks={"PreToolUse": [...]}` |
| System prompt | `systemPrompt` | `system_prompt` |
| Max turns | `maxTurns` | `max_turns` |
| Allowed tools | `allowedTools` | `allowed_tools` |
| MCP servers | `mcpServers` | `mcp_servers` |
| Subagent def | `AgentDefinition` | `AgentDefinition` dataclass |

## Shared Concepts (both languages)

Both SDKs wrap the Claude Code CLI and share these concepts:
- **Hooks**: Pre/PostToolUse, Stop, SubagentStop, Notification, etc.
- **Permissions**: default, acceptEdits, plan, bypassPermissions, etc.
- **MCP Servers**: stdio, HTTP, SSE, SDK (in-process) configurations
- **Subagents**: Delegate tasks to child agents with scoped tools/permissions
- **Structured Outputs**: JSON Schema validation on agent output
- **Sandbox**: Container-based isolation (Docker/Kubernetes)
- **Sessions**: Capture, resume, fork, and mirror transcripts to external storage (`SessionStore`)

For API details, code examples, options tables, and known issues,
read the language-specific reference file.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [xiaolai](https://github.com/xiaolai)
- **Source:** [xiaolai/anthropic-docs](https://github.com/xiaolai/anthropic-docs)
- **License:** MIT
- **Homepage:** https://github.com/xiaolai/autoupdated-anthropic-documentation-knowledge

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-xiaolai-anthropic-docs-claude-agent-sdk
- Seller: https://agentstack.voostack.com/s/xiaolai
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.