Claude ChatLesson 9 of 12

Settings & Toggleable Features

Several Claude Chat features are optional — web search, Deep Research, code execution, and memories. Turn them on or off per chat depending on what you are doing.

Toggleable features overview

Many Claude Chat capabilities are opt-in. You enable them from the chat input toolbar or your account settings. Each conversation can use a different combination — search on for news queries, code execution on for data analysis, all off for a simple writing task.

claude.ai — Chat tools

Toggleable features

Enable or disable capabilities per conversation from the chat input toolbar

Web search

Search the internet for current information

Deep Research

Multi-step research across many sources

Off

Code execution

Run Python and create files in a sandbox

Memory

Save preferences learned from conversations

Toggle features on or off from the chat toolbar — each conversation can use different settings

Web search toggle

Control whether Claude can search the internet during a conversation. When enabled, Claude fetches live information and cites sources. When disabled, answers come only from training data and uploaded files.

When to use

Turn on for current events, live prices, recent news, or time-sensitive facts
Turn off for creative writing, coding help, or tasks that do not need the web
Toggle per conversation — not a permanent account setting

Example: Enable web search, then ask: “What were the major tech product announcements this month?”

Web searchAttachTools

You:

“What is the current inflation rate in the US?”

Searching the web…

Web search active — Claude fetches live data and cites sources

Deep Research mode

A more thorough research mode where Claude plans a research strategy, searches multiple sources, reads and cross-references them, and produces a longer report with multiple sources.

When to use

Comparing options (tools, vendors, approaches) that need multiple sources
Writing research summaries or literature reviews
Questions where a single web search result is not enough
Available on paid plans — slower, but pulls from more sources

Example: Enable Deep Research, then ask: “Compare the top 3 project management tools for remote teams in 2026.”

Deep Research in progress

✓Planning research approach

✓Searching 12 sources

✓Reading and cross-referencing

○Writing detailed report

Topic: “Compare cloud providers for a startup in 2026”

Deep Research runs several steps before producing a longer report

Code execution and file creation

Lets Claude run Python code in a secure sandbox, see the output, and create downloadable files — spreadsheets, charts, and processed datasets — in Anthropic's sandbox rather than on your machine.

When to use

Analysing CSV or Excel data with real calculations
Generating charts or visualisations from data
Testing code snippets and verifying they work before copying
Creating files (CSV, images, PDFs) you can download directly

Example: Upload a sales.csv, enable code execution, and ask: “Group revenue by region and create a summary CSV.”

analysis.py — running…Executed

import pandas as pd
df = pd.read_csv("sales.csv")
print(df.groupby("region")["revenue"].sum())

Output:
region
North   45200
South   38100
West    52400

Created:sales_summary.csvDownload

Code runs in Anthropic's sandbox — output and files are downloadable

Memories from chat history

Claude can learn preferences from your conversations and save them as memories — things like your role, preferred tone, formatting habits, or ongoing projects. Memories persist across chats so you do not repeat yourself.

When to use

When Claude notices a pattern — e.g. you always want bullet points
Saving your role or expertise level for tailored answers
Remembering project context that applies across conversations
You review and approve each memory before it is saved

Example: After several chats where you ask for concise answers, Claude suggests: “Save that you prefer concise responses?” — you click Approve.

Memories — learned from your chats

Preferred toneConcise, professional

RoleSoftware developer

Format preferenceBullet points for lists

Claude suggested saving “I prefer bullet points” from a recent conversation — you approved it.

Memories persist across chats after you approve each one

Managing your settings

Feature	Where to control it	Scope
Web search	Chat input toolbar — Web search label	Per conversation
Deep Research	Chat input toolbar or Research menu	Per conversation
Code execution	Chat input toolbar — Code label	Per conversation
Memories	Settings → Memories	Account-wide (you approve each)

Plan note: Some features — Deep Research, code execution, and extended memory — may require a Pro or Team plan. Free accounts have access to basic web search and limited memory. Check your plan at Settings → Plan on claude.ai.

Before you continue

Web search — turn live internet access on or off per chat.
Deep Research — longer reports pulled from multiple sources.
Code execution — run Python, see output, and download created files.
Memories — Claude learns your preferences with your approval.
Next lesson: formatting and response style.

What's Next

Formatting & Response Style — Markdown output, length, and formats like JSON and bullet lists.

← PREVIOUS

Conversations & History

Formatting & Response Style