Claude Code
roadmap.sh: https://roadmap.sh/claude-code
Suggested path through the Claude Code nodes. Each node links to its lesson when written.
Nodes
Getting started
- what-is-claude-code
- installation-setup
- authentication-api-keys
- first-session-repl-basics
- choosing-a-model
- pricing-token-usage
CLI usage
- interactive-mode
- one-shot-print-mode-p
- piping-input-output
- cli-flags-options
- resuming-continuing-sessions
- managing-context-clear-compact
- image-file-input
- keyboard-shortcuts-vim-mode
Project configuration
- claude-md-project-memory
- nested-imported-memory-files
- user-vs-project-vs-local-memory
- claudeignore-file-scoping
Slash commands
- built-in-slash-commands
- custom-slash-commands
- command-arguments-arguments
- namespaced-project-commands
- frontmatter-allowed-tools
Tools & permissions
- built-in-tools-read-edit-bash-etc
- permission-modes
- allow-deny-rules
- plan-mode
- auto-accept-bypass-modes
Settings
- settings-json-hierarchy
- user-project-settings
- environment-variables
- model-output-configuration
- managed-enterprise-policy-settings
Hooks
- hooks-overview-lifecycle
- pretooluse-posttooluse-hooks
- userpromptsubmit-stop-hooks
- notification-sessionstart-hooks
- writing-debugging-hook-scripts
MCP (Model Context Protocol)
- what-is-mcp
- adding-mcp-servers
- stdio-vs-sse-http-servers
- mcp-tools-resources-prompts
- mcp-authentication-scopes
- building-a-custom-mcp-server
Subagents
- what-are-subagents
- defining-custom-subagents
- subagent-system-prompts-tools
- delegating-tasks-to-subagents
- the-task-tool-parallelism
IDE integrations
Agent SDK
- claude-agent-sdk-overview
- typescript-sdk
- python-sdk
- programmatic-sessions-streaming
- custom-tools-mcp-in-the-sdk
- building-headless-agents
Workflows & automation
- headless-ci-usage
- github-actions-integration
- git-pr-workflows
- multi-step-agentic-tasks
- cost-context-optimization
- security-safe-tool-use
Resources
See resources.md.
Project ideas
- Author a custom slash command plus a PreToolUse hook that lints or guards Bash calls, and prove it fires on a real edit-and-test loop.
- Build a small MCP server (e.g. wrapping an internal API) and wire it into Claude Code, then drive it end-to-end from a session.
- Use the Agent SDK to script a headless agent that triages GitHub issues in CI, posting a summary comment with a custom subagent for code search.