Anthropic Agent SDK Workshop
https://github.com/anthropics/agent-sdk-workshop- Category
- AI Agents
- Rank
- No. 927Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- anthropics
- GitHub
- 132 stars
- Date
About
Anthropic's official workshop for building agents with the Claude Agent SDK — hands-on examples.
What it does
A staged learning environment that reveals agent architecture one capability at a time. You begin with plain chat, then enable tools, specialist workers, and cross-session memory. Later exercises let you assemble scenario agents from prepared components and prompts.
Why it's ranked here
The strongest reason to choose it is instructional clarity. Each stage maps a visible behavior change to a specific SDK primitive, while setup checks, prompt inspection, troubleshooting guidance, and mock data keep attention on agent design. It is less convincing as production validation because integrations and evaluation remain outside the workshop.
What's good
The progression makes abstract concepts observable: tools add lookup, isolated specialists separate research from synthesis, and a lifecycle hook restores saved context. Prepared scenarios cover executive support, customer service, incident investigation, and account risk. A single command checks prerequisites, runs exercises, lists scenarios, shows assembled context, and clears memory.
Tradeoffs
All supplied tools use local mock data, so learners do not encounter authentication, rate limits, changing schemas, or failures from real services. An Anthropic API key and paid model calls are required. Evaluation is manual, with no built-in harness. The guided configuration bypasses tool approval prompts, which the source explicitly distinguishes from a likely production setup.
How to use it well
Use it for a short team workshop or individual lab before implementing a Python agent. Repeat one task while adding capabilities, inspect the assembled context, then adapt a breakout and replace mocks with real integrations. It suits prompt authors and developers learning orchestration. It does not supply production integrations, automated evaluation, or a deployment platform.
Technical notes+
requirements.txt installs claude-agent-sdk>=0.1.41 and python-dotenv>=1.0.0. check_setup.py validates Python, the environment file, imports, and API connectivity. 01-guided-demo/agent.py builds ClaudeAgentOptions, registers in-process MCP servers, allowlists tools, attaches AgentDefinition workers through Task, and installs a UserPromptSubmit hook. 01-guided-demo/memory.py stores up to 50 timestamped notes in JSON and injects them as additionalContext. 01-guided-demo/tools.py serves static JSON-backed research tools. extend/_template/config.py exposes component lists, model, verbosity, turn limit, and data directory settings.
Observed
- Primary language
- Python
- Runtime requirement
- Python 3.10 or newer
- Installation surface
- pip requirements file with Claude Agent SDK and python-dotenv dependencies
- Interface
- Command-line workshop runner for setup checks, demos, breakouts, prompt inspection, and memory reset
- Tool protocol
- Custom tools are bundled as in-process MCP servers
- Platform support
- Direct command usage on Unix-like systems and Python-prefixed commands on Windows
- External requirements
- Anthropic API key required; supplied exercise tools use local mock data
Read from README.md, requirements.txt, docs/FAQ.md, docs/CHEATSHEET.md, docs/TROUBLESHOOTING.md, check_setup.py, 01-guided-demo/agent.py, 01-guided-demo/tools.py, 01-guided-demo/config.py, 01-guided-demo/memory.py, 01-guided-demo/subagents.py, extend/_template/run.py, extend/_template/config.py, 02-breakouts/freeform/run.py, 02-breakouts/00-warmup/run.py.
What it can do
Provide hands-on SDK tutorials
Workshop curriculum and exercises → Interactive coding examples and guided implementations
Demonstrate agent creation workflows
Claude Agent SDK components and APIs → Working agent implementations and code samples
Guide integration of Claude models into applications
Application requirements and use cases → Implementation patterns and integration code
Teach agent conversation management
User interaction patterns and dialogue flows → Conversational agent templates and examples
Show tool and function calling implementation
External APIs and custom functions → Agent configurations with tool integration code
Provide debugging and testing examples
Agent code and behavior scenarios → Testing frameworks and debugging techniques
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.