Settings & Shortcuts Collection: Efficiency Techniques to Master

Introduction: Your perceived speed with Claude changes with "settings" and small tricks

Once you start using Claude every day, what matters more than the model’s performance itself isoperational speed andhow easily you can organize the conversation. Especially as of 2025, Claude is strong at reading long-form text, writing, and assisting with code, but if you use it carelessly, small losses add up—things like "previous instructions getting buried," "the conversation becoming too long," or "not getting the output format you want."

In this article, centered on convenient features that are surprisingly easy to overlook, we’ll compile practical efficiency techniques that work in day-to-day work—up toshortcuts, conversation forks, search, file attachments, locking the output language, leveraging Markdown, and external integrations like Zapier.

First, the settings to lock in: Reduce rework every time

Fix the output language and tone from the start

Claude does a good job picking up context, but as conversations grow longer, the output format can gradually drift. That’s why it’s effective to place astanding rule at the beginning of the chat.

In this conversation, please always follow the rules below.
- Answers must be in Japanese
- Use desu/masu style
- Organize in the order: headings → bullet points → conclusion
- Output in Markdown except for code
- If something is unclear, don’t guess—ask for confirmation

Even for a one-off request, simply placing these rules in the very first turn can significantly reduce the number of revision cycles afterward. If you use them often, it’s convenient to save them in a notes app or a text-expansion tool.

Lock in a “role” per project

Separating chats by purpose is also effective. For example, use them like this.

• For writing text: prioritize readability and remove overly verbose phrasing
• For research and organization: prioritize candidate sources, comparisons, and framing key points
• For code assistance: make differences, error causes, and reproduction steps explicit

Rather than stuffing everything into one conversation, splitting by purpose helps stabilize Claude’s response quality.

Keyboard shortcuts: Faster just by reducing mouse movement

In Claude’s web app, there are differences depending on the environment, but using shortcuts can greatly improve your operation tempo. In particular, it’s worth learningnew chat, send, search, and moving between conversations. Confirm the exact shortcuts in the shortcut list inside the app.

The key point is not to try to memorizeall shortcuts. First, it’s enough to learn only the three: "new chat," "send," and "search." Memorizing the operations you use daily makes it stick the most in the end.

Conversation forks: Branch similar requests to compare

Especially useful in Claude isconversation forks. This is a feature that lets you branch off from a certain point in a conversation and try another pattern. It’s particularly valuable for beginners.

When forks are a good fit

• For the same article outline, create abeginner version and anadvanced version
• Compare a sales email as apolite version versus ashort version
• Test code changes withminimal edits versusfull refactoring

Instead of repeating in one conversation things like "maybe we should try another idea" and "I want to go back to the previous one," it becomes far easier to stay organized when you branch and compare in parallel.

Fork this response and create the following two patterns.
A: Reduce jargon for beginners
B: Expand the steps in detail for practitioners

You can think of forks as afeature that makes trial and error safe. You can improve things while keeping the original draft you don’t want to lose.

Conversation search: Use past logs as “assets”

As you keep using Claude, good past answers tend to get buried. That’s where you should use conversation search. The trick is to leave searchable words in your conversation yourself.

How to make searching easier

• Write the project name at the beginning of the conversation
• Include specific terms like "invoice automation," "recruitment PR," or "LP improvements"
• Explicitly state the final deliverable format (e.g., proposal, SQL, FAQ)

For example, even if you think "the tone of the FAQ we made before was good," you may struggle to rediscover it if the conversation name is vague. Justdesign your conversations with searching in mind can significantly improve reuse efficiency.

Tips for file attachments: Don’t send too much at once

Claude is good at reading files, but if you dump a large number of them carelessly, the intent becomes blurry. The important thing is to sendfiles + instructions + output format together as a set.

Example of a good attachment instruction

Read the attached minutes PDF and create the following.
- Decisions made
- Outstanding items
- To-dos by person
- A 200-character summary
Please output in Markdown with headings.

How to use it effectively

• Clarify the role of each file (specs, minutes, drafts, etc.)
• Write what to extract first
• Specify the final format (table, bullet points, JSON, etc.)
• When multiple files are included, distinguish the “comparison target” and “reference materials”

Also, PDFs with low OCR quality or broken tables can lead to misreading. For important documents, adding the relevant sections as text helps stabilize accuracy.

Leveraging Markdown: Ask for output in a format you can copy as-is

Claude’s output becomes easier to reuse when you assume Markdown. It’s easy to paste into Notion, GitHub, Qiita, internal wikis, etc., and it’s also simple for people to edit later.

Please output in the following format.
## Overview
## Background
## Steps
## Notes
## Next actions
If you need a table, please use a Markdown table.

What’s especially recommended isgiving the heading template upfront. Compared to free generation, it reduces missing information and keeps the order from drifting.

External integrations: Automate “routine work” with Zapier

Claude alone is already quite helpful, but combining it with automation tools like Zapier can have a big impact on reducing routine tasks.

Common integration examples

• Form submission → summarize in Claude → notify on Slack
• Incoming Gmail → categorize contents → record in a spreadsheet
• Upload minutes → extract key points → save to Notion

The important thing is not to build complicated automations from the start. Start with a simple line: "receive," "summarize," and "transcribe," and you’ll be less likely to fail.

Example of a prompt to send from Zapier to Claude

Read the following inquiry and return it as JSON.
{
  "category": "invoices/bugs/how-to/other",
  "urgency": "high/medium/low",
  "summary": "Within 100 characters",
  "replyDraft": "A polite first-draft reply"
}

When you requeststructured output like this, subsequent automated processing becomes more stable.

Recommended operating patterns: Easy to copy even for beginners

1. Fix “language, tone, and format” at the beginning of the conversation
2. Split chats by purpose
3. Branch alternative ideas and compare
4. When attaching files, instruct the extraction items and output format together
5. Receive the final deliverable as Markdown or JSON
6. Consider automating repetitive work with Zapier or similar tools

Even with just these six points, yourrepeatable results as a user improve far more than Claude’s “smartness” alone.

Conclusion: Claude gets faster when you’re the one who set it up

The key to getting the most out of Claude isn’t memorizing difficult prompts. Rather, it’s more effective tofix the settings you use often, organize the conversation, and align the output format. Speed up operations with shortcuts, organize trial and error with forks, increase reusability with search and Markdown, and automate routine work with external integrations. Building on these steps steadily boosts your daily productivity.

For the final installment, we encourage you to introduce at least one thing from today. The recommended first step is tofix it every time: “Answer in Japanese using Markdown”. It’s a small setting, but it works far better than you might expect.