Claude Code Tips: A 10-Minute Shortcut Playbook
Start with one real bug, not a repository-wide rewrite. This Claude Code quick-start guide gives you a repeatable workflow: establish project rules, scope the task, review a plan, then verify the patch. Keep the commands below beside your terminal.
1. Teach Claude Your Project Once
Open your repository, launch claude, and run /init to generate a starter CLAUDE.md. Review it before committing: replace guessed commands with the exact build, lint, and test commands your team uses. Include architectural boundaries, such as where API handlers belong. Claude reads these instructions across sessions, so keep them concise and concrete. They are guidance, not security controls. ([code.claude.com](https://code.claude.com/docs/en/memory))
2. Give Every Task a Finish Line
Try this prompt: Investigate the expired-token failure in @src/auth/session.ts. Add a regression test, preserve the public API, and avoid unrelated refactors. The @ shortcut opens file-path autocomplete, letting you point Claude toward relevant code instead of pasting entire files. ([code.claude.com](https://code.claude.com/docs/en/interactive-mode))
For changes spanning several files, press Shift+Tab until Plan Mode appears. Ask for affected files, proposed changes, and validation steps before approving implementation. You can also start directly with claude --permission-mode plan. ([code.claude.com](https://code.claude.com/docs/en/common-workflows))
3. Memorize These Keyboard Shortcuts
Bindings can vary by terminal; press ? on an empty prompt to see shortcut help. These are useful starting points: ([code.claude.com](https://code.claude.com/docs/en/interactive-mode))
- Esc: interrupt the current response or tool call when Claude heads in the wrong direction.
- Ctrl+R: search prompt history instead of rewriting a useful instruction.
- Ctrl+G: open your prompt in your default text editor for longer specifications.
- Ctrl+J: insert a newline without submitting, useful for separating requirements from acceptance criteria.
- Ctrl+O: open the detailed transcript to inspect tool activity. ([code.claude.com](https://code.claude.com/docs/en/interactive-mode))
4. Reset Context Without Losing Project Rules
Use /context to inspect what occupies the context window. When switching to an unrelated task, run /clear; it starts fresh while retaining project instructions. Mid-task, use /compact Preserve the failing test, root-cause hypothesis, and remaining steps. Compaction summarizes conversation history rather than keeping every detail verbatim, so save critical decisions in project files. ([support.claude.com](https://support.claude.com/en/articles/14553413-claude-code-cheatsheet))
5. Recover Safely and Resume Quickly
Use /rewind to return to an earlier checkpoint. With an empty prompt, pressing Esc twice also opens the rewind menu; with a draft present, it clears the draft instead. ([code.claude.com](https://code.claude.com/docs/en/interactive-mode))
Do not treat rewind as Git. It does not undo file changes made through Bash commands, and external changes are not reliably captured. Keep a clean branch and review your diff before committing. ([code.claude.com](https://code.claude.com/docs/en/checkpointing))
Returning after a meeting? Run claude --continue in the same directory to resume its latest session, or claude --resume to choose a previous conversation. ([code.claude.com](https://code.claude.com/docs/en/common-workflows))
6. Keep API Costs Predictable With 59API
59API is a practical low-cost option for developers who prefer prepaid, pay-as-you-go access over another monthly subscription. It advertises official-quality model capability without downgrades, compatibility with Claude Code, Codex, and OpenAI SDKs, plus a referral rebate. Compare the current price for your chosen model rather than assuming every request has the same discount. ([59api.com](https://59api.com/?lang=en))
To connect an existing Claude Code installation, create a 59API key and add credit. In the shell that launches Claude, set ANTHROPIC_BASE_URL to 59API's published API endpoint and ANTHROPIC_AUTH_TOKEN to your key, then run claude. Keep credentials out of committed project files. If flexible billing fits your workflow, consider signing up for 59API and testing one small task first. ([59api.com](https://59api.com/?lang=en))
Use /model to inspect or switch models. Check token usage with /cost, but reconcile actual charges against your relay account: Claude Code's dollar figures are estimates and may use list prices rather than your provider's rates. ([support.claude.com](https://support.claude.com/en/articles/14553413-claude-code-cheatsheet))
Your Next Five Minutes
Pick one failing test. Supply its reproduction steps, approve a focused plan, and ask Claude to report the exact validation commands and results. Review the diff yourself. The goal is not fewer keystrokes at any cost; it is a small, understandable patch you can confidently ship.