Open claude.ai and type a message into the box at the center of the screen to begin.
Example first message:"I'm planning a 3-day trip to Chicago in October. Can you suggesta rough itinerary that mixes food, museums, and walkable neighborhoods?"
There's no setup step between signing in and talking to Claude - the message box is the whole interface.
A new, blank conversation starts with zero context, so the more relevant detail you include, the better the first reply.
Claude replies in the same window, and you can keep replying to build on what it just said.
You don't need to phrase things as a "command" - plain, natural sentences work best.
The composer is the text box at the bottom (or center, before your first message) of every conversation.
Composer basics:- Type your message and press Enter (or click the send arrow) to send it.- Shift+Enter adds a new line without sending.- The paperclip/attach icon lets you add a file or image to the same message.
The composer is where every turn of the conversation starts, whether it's your first message or your tenth.
You can paste large blocks of text directly into it - long emails, articles, or notes - and Claude will treat that as part of your message.
While Claude is replying, the send control becomes a stop control so you can interrupt a response that's heading the wrong way.
Edited or resent messages replace that point in the conversation going forward, which is useful when you want to correct course.
A model picker near the composer lets you choose which Claude model handles your message.
Model picker (typical Free/Pro options):- Claude Sonnet 5 - the default, balanced for most everyday tasks.- Claude Opus 4.8 - stronger reasoning for harder or higher-stakes tasks.- Claude Haiku 4.5 - fastest, best for quick or simple requests.- Claude Fable 5 - available depending on your plan and use case.
Claude Sonnet 5 is the default for Free and Pro accounts, and it's a reasonable choice for most day-to-day questions.
Which models you can pick from depends on your plan tier - Free, Pro, Team, and Enterprise differ in model access as well as usage limits.
You can switch models mid-conversation; the new model picks up the existing context and continues from there.
There's no penalty for experimenting - if a reply from one model doesn't fit your needs, try the same message with another.
A "New chat" control (usually in the sidebar) clears the slate for an unrelated task.
When to start new:- You're switching to a completely different topic or project.- The current conversation has grown long and lost focus.- You want a clean first answer without earlier back-and-forth influencing it.
A new conversation carries no memory of a previous one, so restate any background Claude still needs to know.
Starting fresh is not "losing progress" - your old conversation is still saved and searchable in your history.
This is the single most useful habit for keeping replies sharp: one conversation per task, rather than one giant running thread.
If you're unsure whether to continue or restart, ask yourself whether the new question is a follow-up to the current topic or a new one.
The sidebar lists your recent conversations, and a search field finds older ones by title or content.
Sidebar basics:- Recent conversations appear in reverse chronological order.- Click the search icon and type a keyword to find an older conversation.- Click any conversation in the list to reopen it exactly where you left off.
Every conversation you have is saved automatically - there's no manual "save" step.
Reopening a past conversation restores its full context, so you can pick a task back up days later.
Search matches against conversation content, not just titles, so you can find a thread even if you don't remember what you named it.
This is also how you sync work across devices - the same history appears in the desktop and mobile apps once you're signed in.
If a response isn't quite right, you can edit your message or ask Claude to try again instead of starting over.
Two ways to redirect a reply:- Edit your last message and resend it - Claude replaces its answer based on the new wording.- Click "Regenerate" (or ask directly, e.g. "try a shorter version") to get another attempt at the same question without changing what you asked.
Editing a message effectively rewinds the conversation to that point, so anything after it is replaced.
Regenerating keeps your original message intact and just asks for a different attempt at the answer.
This is often faster than opening a new conversation when the issue is tone, length, or format rather than the underlying question.
Combining this with a specific note - "shorter," "more formal," "skip the bullet points" - usually gets a better second attempt than a bare retry.
Switching models deliberately, rather than always using the default, gets better results for harder or simpler jobs.
A practical pattern:1. Start a quick, low-stakes question on Claude Haiku 4.5 for speed.2. Switch to Claude Sonnet 5 (the default) for typical everyday work.3. Switch to Claude Opus 4.8 for a task that needs deeper reasoning, like reviewing a complex document or working through a hard problem.
There's no separate interface to learn - switching models just changes which one answers your next message in the same conversation.
Faster models are a good fit for simple, high-volume requests; stronger models are worth the extra time for complex or high-stakes ones.
Model access varies by plan tier, so which of these options you see depends on whether you're on Free, Pro, Team, or Enterprise.
This same model choice appears consistently across claude.ai, the desktop app, and mobile - it isn't a web-only setting.
Free and Pro plans have usage limits that reset periodically; watching for them helps you plan heavier work sessions.
Signs it may be time to look at a plan upgrade:- You're regularly hitting a "usage limit reached" message before your work is done.- You need a stronger model than your current plan allows for a specific task.- You're coordinating with a team and need shared or admin features rather than just an individual account.
Free and Pro accounts have usage limits that reset on a rolling basis; Team and Enterprise plans are built around higher, more predictable limits and admin controls.
Hitting a limit doesn't lose any conversation history - it just pauses new messages until the limit resets or you upgrade.
If you're evaluating Claude for a team rather than yourself, Team and Enterprise tiers add collaboration and admin features that Free and Pro don't include.
This is a plan-level setting, not something you configure per conversation - check your account or plan page to see current limits and options.
Stack versions: Written against the Claude model lineup current as of ~June 2026 - Claude Fable 5, Claude Opus 4.8, Claude Sonnet 5 (the default), and Claude Haiku 4.5. Model names, pricing, and product features move quickly - verify current specifics at platform.claude.com/docs before relying on them.
Reviewed by Chris St. John·Last updated Jul 18, 2026