Claude SDK Quest

Tracks

1. 01🧭SDK Map and Current Claude Surface

   0/8 lessons

   Three Claude SDKs, model picker, OAuth landscape, and provider boundaries

   Separate Anthropic Client SDKs from the renamed Claude Agent SDK and the Claude Code CLI. Pick the right surface for the job and frame every call against the current Claude model lineup.

   Lesson list (8)

   • 01
     Three Claude SDKs Are Not One Thing~18 min · sdk-map, client-sdk, agent-sdk, claude-code-cli
   • 02
     The Current Claude Lineup and How to Pick~16 min · models, opus, sonnet, haiku, model-id
   • 03
     Auth Paths: API Keys, Console, OAuth, and Bedrock/Vertex~18 min · auth, api-key, oauth, bedrock, vertex
   • 04
     Regions, Quota, and Rate Limit Reality~14 min · rate-limits, quota, regions, throughput
   • 05
     Docs as Source of Truth, Models as Suspect~12 min · docs, knowledge-cutoff, verification
   • 06
     Security and Privacy at the SDK Boundary~14 min · security, privacy, secrets, logging
   • 07
     Ecosystem at a Glance: MCP, Tools, and Cookbooks~12 min · mcp, ecosystem, cookbook
   • 08
     How cwkPippa Uses All Three Surfaces~16 min · cwkpippa, self-reference, architecture

   ✓Quiz · 4 questions→

2. 02✉️Messages API: The Surface Underneath Everything

   0/6 lessons

   Request shape, system prompts, multi-turn, streaming, structured output, and the cost knobs

   The Messages API is the wire format under every Anthropic client. Master the request shape, the response shape, and the four levers (system / messages / max_tokens / stop_sequences) and the rest of the SDK becomes a thin convenience wrapper.

   Lesson list (6)

   • 01
     The Anatomy of a Messages Request~16 min · messages-api, request, system, max-tokens
   • 02
     System Prompts: The Largest Single Lever~16 min · system-prompt, instructions, persona
   • 03
     Multi-Turn Conversations and Memory~14 min · multi-turn, memory, history
   • 04
     Streaming: Time-to-First-Token Matters~16 min · streaming, sse, events
   • 05
     Stop Sequences, Stop Reasons, and Forcing Structure~14 min · stop-sequences, stop-reason, json, structured-output
   • 06
     Errors, Retries, and the Idempotency Story~14 min · errors, retries, idempotency, 5xx

   ✓Quiz · 4 questions→

3. 03🐍The Python SDK Deep Dive

   0/6 lessons

   Sync, async, streaming, tool loops, prompt caching, files, and batches in idiomatic Python

   Master the anthropic Python SDK as a working tool — sync vs async clients, streaming patterns, tool loops, prompt caching, the Files API, and the Batch API. Every example is runnable; every pattern survives in production.

   Lesson list (6)

   • 01
     Install, Pin, and Verify the SDK~12 min · install, pinning, version
   • 02
     AsyncAnthropic in FastAPI and Long-Running Workers~16 min · async, fastapi, concurrency
   • 03
     Streaming Helpers, Final Messages, and Usage~14 min · streaming, stream-helper, usage
   • 04
     Tool Loops in Python: From One Round to Many~18 min · tool-use, tool-loop, function-calling
   • 05
     Prompt Caching for Real Dollar Savings~16 min · prompt-caching, cache-control, ephemeral
   • 06
     Files API and Batch API: Async-First Workflows~16 min · files-api, batch-api, async-workflows

   ✓Quiz · 4 questions→

4. 04🟦The TypeScript SDK in Depth

   0/6 lessons

   Strict types, streaming on Node and Edge, tool loops, and Next.js patterns

   Use the @anthropic-ai/sdk TypeScript client like a pro: strict types, streaming on Node and Edge runtimes, tool loops, and Next.js Route Handlers. Same API surface as Python, different idioms.

   Lesson list (6)

   • 01
     Install, Types, and the One Client~14 min · install, types, esm, cjs
   • 02
     Streaming on Node and Next.js~16 min · streaming, next-js, route-handler
   • 03
     Typed Tool Loops in TypeScript~18 min · tool-use, typed, registry
   • 04
     Typed Errors and Retry Posture~12 min · errors, retries, discriminated
   • 05
     Next.js + Claude: Auth, Edge, and Server Components~16 min · next-js, edge, server-components, rsc
   • 06
     Deno, Bun, and Cross-Runtime Reality~12 min · deno, bun, edge, cross-runtime

   ✓Quiz · 4 questions→

5. 05🛠️Tool Use: Custom, Built-in, and Server Tools

   0/6 lessons

   From single function calls to parallel dispatch, server-side tools, code execution, and computer use

   Tool use is the lever that turns Claude from a chatbot into a system component. Master custom tools with JSON Schema, the parallel-tool pattern, server tools (web search, code execution, computer use), and the always-uncomfortable failure modes.

   Lesson list (6)

   • 01
     Tool Definitions and JSON Schema Discipline~16 min · json-schema, input-schema, validation
   • 02
     Parallel Tool Calls in One Round~14 min · parallel, tool-use, concurrency
   • 03
     Tool Errors, Retries, and Telling the Model About Them~14 min · errors, retries, tool-result
   • 04
     Server Tools: Web Search, Code Execution, Computer Use~16 min · server-tools, web-search, code-execution, computer-use
   • 05
     Tool Runner: A Reusable Loop Abstraction~14 min · tool-runner, abstraction, library
   • 06
     Tool Security: Permissions, Sandboxes, and Side Effects~14 min · security, permissions, side-effects

   ✓Quiz · 4 questions→

6. 06🗂️Context Windows, Vision, PDFs, and Documents

   0/6 lessons

   Long context budgets, multimodal inputs, the Files API, and right-sizing your context

   Master Claude's input surface — long context windows, image and PDF inputs, the Files API, RAG patterns, and the discipline of putting only what matters into the model.

   Lesson list (6)

   • 01
     Context Windows: Long Is Not Free~14 min · context-window, tokens, budget
   • 02
     Vision: Sending Images to Claude~14 min · vision, images, multimodal
   • 03
     PDF Input: Two Paths and When to Use Each~14 min · pdf, document, files-api
   • 04
     RAG vs Stuffing: Pick by Lookup Frequency~14 min · rag, retrieval, embeddings
   • 05
     Summarization, Memory Compression, and Drift~12 min · summarization, memory, compression
   • 06
     Context as Architecture~14 min · architecture, context-engine, self-reference

   ✓Quiz · 4 questions→

7. 07📈Scale: Cost, Reliability, and Observability

   0/6 lessons

   Make a Claude-powered system that survives traffic, surprises, and finance review

   Take a working Claude integration to production scale: per-token cost discipline, deep prompt caching, batch workflows, retry posture, observability, and the eval discipline that keeps regressions out.

   Lesson list (6)

   • 01
     Anatomy of a Claude Bill~14 min · cost, billing, tokens
   • 02
     Prompt Caching Strategy in Depth~16 min · prompt-caching, breakpoints, ttl
   • 03
     Batch API Workflows in Production~14 min · batch-api, async, evaluation
   • 04
     Retry Posture and Reliability Patterns~14 min · retries, reliability, circuit-breaker
   • 05
     Observability for LLM Workloads~14 min · observability, tracing, metrics
   • 06
     Eval Discipline: Catch Regressions Before Users Do~14 min · eval, regression, quality

   ✓Quiz · 4 questions→

8. 08🤖Claude Agent SDK Core

   0/8 lessons

   Renamed from Claude Code SDK: install, ClaudeAgentOptions, query vs ClaudeSDKClient, sessions, MCP, hooks, slash commands

   The Claude Agent SDK gives you Claude Code's autonomous harness — filesystem tools, command execution, persistent sessions, MCP integration, and hooks — as a programmable library. This is the SDK that built cwkPippa's Claude brain.

   Lesson list (8)

   • 01
     Renamed: Claude Code SDK Is Now Claude Agent SDK~14 min · rename, install, claude-agent-sdk, migration
   • 02
     query() vs ClaudeSDKClient: Pick by Lifecycle~16 min · query, ClaudeSDKClient, lifecycle
   • 03
     ClaudeAgentOptions: Every Knob That Matters~18 min · options, configuration, ClaudeAgentOptions
   • 04
     The Default Tool Set: What You Get for Free~14 min · tools, default-tools, filesystem, bash
   • 05
     MCP Servers in the Agent SDK~16 min · mcp, model-context-protocol, servers
   • 06
     Hooks: Pre and Post Tool Use~16 min · hooks, PreToolUse, PostToolUse
   • 07
     Sessions, Resume, and JSONL Truth~16 min · sessions, resume, jsonl, persistence
   • 08
     Sub-Agents, Slash Commands, and Skills~14 min · subagents, slash-commands, skills

   ✓Quiz · 4 questions→

9. 09🛡️Agent Governance: Permissions, Security, Audit

   0/5 lessons

   Make an autonomous agent safe to run in production — permission modes, sandboxes, audit trails, and prompt injection defense

   An autonomous Claude agent has authority to read files, run shell, and call APIs. Governance is what keeps that authority bounded: permission modes, sandboxes, prompt-injection defense, audit trails, and the discipline of saying no.

   Lesson list (5)

   • 01
     Permission Modes in Practice~14 min · permissions, permission-mode, safety
   • 02
     Sandboxing: cwd, env, and Network~14 min · sandbox, cwd, environment, network
   • 03
     Prompt Injection Defense~16 min · prompt-injection, security, isolation
   • 04
     Audit Trails for LLM Actions~12 min · audit, logging, compliance
   • 05
     Testing Agent Behavior~14 min · testing, evals, regression

   ✓Quiz · 4 questions→

10. 10🏗️cwkPippa-Grade Production Patterns

    0/7 lessons

    Provider port (or not), fallback chains, JSONL ground truth, persistent subprocess pools, OAuth, deployment, post-mortems

    The end-game patterns that turn a working Claude integration into something you would bet a product on: provider boundaries, fallback chains, durable JSONL truth, persistent subprocess pools, OAuth lifecycle, and the post-mortem discipline.

    Lesson list (7)

    • 01
      Provider Port — When and Where~16 min · provider-port, abstraction, architecture
    • 02
      Fallback Chains and Degraded Modes~14 min · fallback, degraded-mode, reliability
    • 03
      JSONL Ground Truth, Derived Mirrors~16 min · jsonl, persistence, rebuild
    • 04
      Persistent Subprocess Pools (Agent SDK)~14 min · subprocess, pool, agent-sdk
    • 05
      OAuth Lifecycle in Production~12 min · oauth, auth, tokens
    • 06
      Deployment, Monitoring, and Restart Discipline~14 min · deployment, monitoring, restart, launchd
    • 07
      Post-Mortems and the History Log Habit~12 min · post-mortem, history, learning

    ✓Quiz · 4 questions→