站内导读
定位:Anthropic 维护的示例项目集,面向用 Claude API 构建应用的开发者,按客服、数据分析、桌面控制等场景收录 18 个可改造起点。
核心能力
- 18 个示例各自成独立子目录,自带 README 与安装说明,进入对应目录单独装依赖运行。
- 桌面控制示例支持 computer_toolset_20260801,动作由模型按名调用。
- 两个最小示例分别基于 CDP 单标签页与 VNC 的 RFB 协议。
- 日程简报用 ant apply 以文件定义部署,含两处记忆存储、vault 与 cron。
适用人群:适合持 Claude API key、能按子项目装依赖的开发者;示例含 Python、TypeScript,macOS 实现建议放虚拟机。
注意:两个 toolset 最小示例只展示接口,不是生产代码。
位置づけ:Anthropic が保守するサンプルプロジェクト集で、Claude API でアプリを作る開発者に向け、カスタマーサポート、データ分析、デスクトップ制御などの場面ごとに、改造の起点になる 18 件のサンプルを収録しています。
主な機能
- 18 件のサンプルはそれぞれ独立したサブディレクトリで、README とインストール手順を備え、ディレクトリごとに依存を入れて実行します。
- デスクトップ制御のサンプルは computer_toolset_20260801 に対応し、動作はモデルが名前で呼び出します。
- 2 件の最小サンプルは、それぞれ CDP の単一タブと VNC の RFB プロトコルを基盤にしています。
- スケジュールブリーフィングは ant apply でファイル定義のままデプロイでき、2 つのメモリストレージ、vault、cron を含みます。
対象ユーザー:Claude API キーを持ち、サブプロジェクトごとに依存を入れられる開発者に向いています。サンプルは Python と TypeScript で、macOS 向けの実装は仮想マシン上での実行が推奨されます。
注意:2 件の toolset 最小サンプルはインターフェースの提示が目的で、本番コードではありません。
정의: Anthropic이 유지보수하는 예제 프로젝트 모음으로, Claude API로 앱을 만드는 개발자를 위해 고객 지원, 데이터 분석, 데스크톱 제어 등의 상황별로 고쳐 쓸 수 있는 출발점 18개를 수록했습니다.
핵심 기능
- 예제 18개는 각각 독립된 하위 디렉터리로, README와 설치 설명을 갖추고 있어 디렉터리별로 의존성을 설치해 실행합니다.
- 데스크톱 제어 예제는 computer_toolset_20260801을 지원하며, 동작은 모델이 이름으로 호출합니다.
- 최소 예제 두 개는 각각 CDP 단일 탭과 VNC의 RFB 프로토콜을 기반으로 합니다.
- 일정 브리핑은 ant apply로 파일 정의 그대로 배포하며, 메모리 저장소 두 곳, vault, cron을 포함합니다.
대상 사용자: Claude API 키를 갖고 하위 프로젝트별로 의존성을 설치할 수 있는 개발자에게 적합합니다. 예제는 Python과 TypeScript로 되어 있고, macOS 구현은 가상 머신에서 실행하는 것이 권장됩니다.
참고: toolset 최소 예제 두 개는 인터페이스를 보여 주는 용도이며 프로덕션 코드가 아닙니다.
What it is: A collection of sample projects maintained by Anthropic for developers building apps with the Claude API — 18 adaptable starting points organized by scenario such as customer support, data analysis, and desktop control.
Key capabilities
- Each of the 18 samples is a standalone subdirectory with its own README and setup instructions; you install dependencies and run it inside that directory.
- The desktop-control sample supports computer_toolset_20260801, with actions invoked by the model by name.
- Two minimal samples are based on a single CDP tab and on VNC's RFB protocol respectively.
- The schedule briefing deploys via ant apply from file definitions and includes two memory stores, a vault, and cron.
Who it is for: Developers who hold a Claude API key and can install dependencies per subproject. Samples are in Python and TypeScript; the macOS implementation is best run in a virtual machine.
Caveats: The two minimal toolset samples only demonstrate the interface — they are not production code.
README · 详细文档
来源 · 查看原文
Claude Quickstarts
Claude Quickstarts is a collection of projects designed to help developers quickly get started with building applications using the Claude API. Each quickstart provides a foundation that you can easily build upon and customize for your specific needs.
Getting Started
To use these quickstarts, you'll need an Claude API key. If you don't have one yet, you can sign up for free at console.anthropic.com.
Available Quickstarts
Customer Support Agent
A customer support agent powered by Claude. This project demonstrates how to leverage Claude's natural language understanding and generation capabilities to create an AI-assisted customer support system with access to a knowledge base.
Go to Customer Support Agent Quickstart
Financial Data Analyst
A financial data analyst powered by Claude. This project demonstrates how to leverage Claude's capabilities with interactive data visualization to analyze financial data via chat.
Go to Financial Data Analyst Quickstart
Computer Use Demo
An environment and tools that Claude can use to control a desktop computer. This project demonstrates how to leverage the computer use capabilities of Claude, including support for the latest computer_toolset_20260801 toolset shape, where each computer action is a member tool the model calls by name.
Go to Computer Use Demo Quickstart
Computer Use Best Practices
A pedagogical, native-macOS reference implementation of a computer-use agent. Unlike the containerized Computer Use Demo, this project runs directly against the macOS desktop (run it in a VM!) and demonstrates common patterns for building more reliable and cost-effective computer-use agents: explicit tool definitions, correct image sizing and pruning, prompt caching, server-side compaction, batched tool calls, a sandboxed shell, and trajectory recording. It pairs with Anthropic's computer-use best-practices guide.
Go to Computer Use Best Practices Quickstart
Browser Use Demo
A complete reference implementation for browser automation powered by Claude. This project demonstrates how to give Claude the ability to navigate websites, inspect and interact with DOM elements, extract content, and fill forms using a custom Playwright-backed browser tool.
Go to Browser Use Demo Quickstart
Browser Toolset Quickstart
The minimal CDP example of Claude's browser toolset (browser_toolset_20260801), in Python and TypeScript. It has a small driver for one Chromium tab over the Chrome DevTools Protocol, a script that runs the driver with the tool runner, and a script that exercises it without a model. It shows the interface and is not production code.
Go to Browser Toolset Quickstart
Computer Toolset Quickstart
The minimal VNC example of Claude's computer toolset (computer_toolset_20260801), in Python and TypeScript. It has a small driver for one desktop over VNC (the RFB protocol, no VNC library), a script that runs the driver with the tool runner, and a script that exercises it without a model. It shows the interface and is not production code.
Go to Computer Toolset Quickstart
Autonomous Coding Agent
An autonomous coding agent powered by the Claude Agent SDK. This project demonstrates a two-agent pattern (initializer + coding agent) that can build complete applications over multiple sessions, with progress persisted via git and a feature list that the agent works through incrementally.
Go to Autonomous Coding Agent Quickstart
Managed Agents: Chat SDK
A browser chat app powered by Claude Managed Agents and Vercel's Chat SDK. This project demonstrates how to pair a persistent Managed Agents session per conversation with a chat surface: a research analyst researches with web search and streams its brief token by token while a live feed shows the tool calls. The same handler runs on Slack, Teams, Discord, Telegram, or WhatsApp by swapping the Chat SDK adapter.
Go to Managed Agents Chat SDK Quickstart
Managed Agents with CopilotKit and AG-UI
A personal finance assistant chat app built on a Claude Managed Agent and rendered with CopilotKit. This project demonstrates how to bridge a managed-agent session to the AG-UI protocol, stream replies token by token with event deltas, and render agent tool calls as interactive generative UI components in the chat.
Go to Managed Agents with CopilotKit Quickstart
Managed Agents: Daily Brief
A scheduled agent that reads your Slack channels and GitHub pull requests each weekday and posts one short brief to Slack, built on Claude Managed Agents. This project demonstrates a deployment defined entirely as files applied with ant apply (agent, environment, two memory stores, vault, cron deployment), a read-only preferences store next to a read-write state store, and run steps designed to fail well unattended: per-source bookmarks, a ledger against repeats, and a source that could not be read reported as unreadable rather than as a quiet day.
Go to Managed Agents Daily Brief Quickstart
Managed Agents: Knowledge Wiki
A deal-room knowledge wiki built with Claude Managed Agents. This project demonstrates how to distill a document corpus once into a versioned memory-store knowledge wiki — using one dynamic workflow run that extracts, resolves, and consolidates — then answer repeated analyst questions from the wiki with provenance on every fact and a lower per-question cost than raw-document search. The worked example is a real M&A data room fetched from public SEC EDGAR filings.
Go to Managed Agents Knowledge Wiki Quickstart
Managed Agents: Linear
An agent you can @mention or assign in Linear, backed by a Claude Managed Agent. This project demonstrates a stateless webhook bridge on Linear's Agent Platform: an AgentSessionEvent creates a Managed Agents session with the Linear session and organization IDs stored in session metadata, and the session.status_idled webhook reads that metadata back to post the reply as a comment. It installs through Linear OAuth with actor=app.
Go to Managed Agents Linear Quickstart
Managed Agents: MCP Server (TypeScript)
An MCP server that exposes the Claude Managed Agents Sessions API as nine tools. This project demonstrates how to let Claude Desktop, Claude Code, or claude.ai drive the agents already in your workspace: list them, start a session, send a message, and wait for the reply, over stdio or Streamable HTTP. One tool, wait_for_idle, turns the session's event stream into a single request/response call, and an agent allowlist limits what a bearer-token holder can reach.
Go to Managed Agents MCP Server (TypeScript) Quickstart
Managed Agents: Road Trip Planner
A national-park road trip planner built directly on a Claude Managed Agents session, in Next.js with no chat framework and no database. This project demonstrates four API features on one screen: token-by-token streaming with event_deltas on the session event stream, vault credentials injected at a specific request location (injection_location), a per-session model override with agent_with_overrides, and a multiagent coordinator that hands its draft to a reviewer agent in a session thread.
Go to Managed Agents Road Trip Planner Quickstart
Managed Agents: Sentry
A scheduled Sentry triage agent built on Claude Managed Agents. This project demonstrates a deployment that starts a session on a cron schedule with no host process, Sentry's Agent Plugin for guided setup, and a refreshable MCP OAuth credential persisted in a vault for issue triage and Seer root-cause analysis.
Go to Managed Agents Sentry Quickstart
Managed Agents: Sentry Autofix
A production error fixed by pull request, with Claude Managed Agents and the Sentry MCP server. This project demonstrates a webhook-driven session per Sentry issue: the agent reads the stack trace and Seer's root cause analysis, reproduces the bug with a failing test in a mounted GitHub repository, and pushes a branch, and the host opens the pull request through a custom tool so a person reviews every fix. Setup runs in Claude Code with Sentry's Agent Plugin. Credentials stay in a vault outside the sandbox, the sandbox has no route to the open internet, and each session has a hard spend cap.
Go to Managed Agents Sentry Autofix Quickstart
Managed Agents: Slack
A Slack bot backed by a Claude Managed Agent. This project demonstrates a stateless webhook bridge: an @mention creates a Managed Agents session with the channel and thread stored in session metadata, and the session.status_idled webhook reads that metadata back to post the reply in-thread. There is no database and no long-lived connection.
Go to Managed Agents Slack Quickstart
General Usage
Each quickstart project comes with its own README and setup instructions. Generally, you'll follow these steps:
- Clone this repository
- Navigate to the specific quickstart directory
- Install the required dependencies
- Set up your Claude API key as an environment variable
- Run the quickstart application
Explore Further
To deepen your understanding of working with Claude and the Claude API, check out these resources:
- Claude API Documentation
- Claude Cookbooks - A collection of code snippets and guides for common tasks
- Claude API Fundamentals Course
Contributing
We welcome contributions to the Claude Quickstarts repository! If you have ideas for new quickstart projects or improvements to existing ones, please open an issue or submit a pull request.
Community and Support
- Join our Anthropic Discord community for discussions and support
- Check out the Anthropic support documentation for additional help
License
This project is licensed under the MIT License - see the LICENSE file for details.