Workbranch › Guides

Guide

Claude says the conversation is too long: how to continue in a new chat

You've built up hours of context in one Claude chat and now it says the message would exceed the length limit. You don't have to start from zero. This guide shows how to move the important parts into a new chat and keep the next one from hitting the wall.

Updated 20 October 2026

Quick answer: every Claude chat has a maximum size, called the context window. When you see Your message will exceed the length limit for this chat, ask Claude for a short handoff summary of the chat, then paste it into a new conversation, ideally inside a project that already holds your files and instructions. If the problem is one very large message or file, split it up instead.

The steps below were checked against Anthropic's help center as of October 2026. Menu names may change in later versions of the app.

Why Claude says the chat is too long

The context window is everything Claude can keep in view in one conversation: your messages, its replies, and any files you attached. Part of it is always held back for Claude's answer, so the real maximum is a bit smaller than the advertised size. The size depends on the model you use.

On paid plans with code execution turned on, Claude manages this for you. As a chat gets close to the limit, it summarizes the earlier messages and carries on, and you may see a note that Claude is organizing its thoughts. Your full history stays in the chat. That is why most paid users rarely see the error at all. It still appears when:

The automatic summarizing isn't free: long chats that keep being summarized use more of your usage limit, so Anthropic suggests starting a new chat when you're getting close to that limit anyway.

Option 1: fix a message that's too big

If the error appears right after you paste a long document or attach several files, the chat itself may be fine. Anthropic's suggestions:

Option 2: move to a new chat with a handoff summary

When the conversation itself is full, start over without losing the thread:

  1. If the old chat still accepts a short message, ask for a handoff. For example: Summarize this conversation for a fresh chat: the goal, what we decided and why, what's done, what's still open, and the exact next step. Keep it under 300 words.
  2. Read the summary and fix anything that is wrong or missing. It's your new starting point, so correct it now.
  3. If the old chat won't take any more messages, scroll back and copy the key decisions and the latest version of whatever you were working on into a note yourself.
  4. Start a new chat, paste the summary as your first message, and add the one thing you want to do next.
  5. Attach only the files the next step needs, not everything from the old chat.

Option 3: use a project so context carries over

Projects are available on every Claude plan, including free (up to five projects on free). A project keeps its own files, instructions and chat history, so every new chat inside it starts with that background instead of an empty page.

  1. Create a project for the work and upload the documents you keep referring to into its knowledge.
  2. Write short project instructions: the general background and the role you want Claude to take.
  3. Start each new chat inside the project and paste your handoff summary there.

Project files help with length too. Claude pulls in only the relevant parts of project knowledge, and on paid plans it switches to retrieval automatically when the knowledge grows large, which Anthropic says expands capacity up to ten times. Full setup steps are in How to use Claude Projects.

Option 4: let Claude look up the old chat

On Pro, Max, Team and Enterprise, Claude can search your past chats. Check that Search and reference chats is on under Settings → Memory, then ask in the new chat, for example Let's continue where we left off on the pricing page. Inside a project, the search covers only that project's chats. More detail is in how to use Claude memory and search past chats. A summary you wrote yourself is still more reliable than a search, because you decide what goes in.

Good to know

If it doesn't work

Save the open tasks before you start over

A handoff summary lives in one chat. If you want the open items to survive more than one restart, Workbranch for Claude can hold them: in the long chat, say save this to Workbranch, and Claude files its tasks, decisions, open questions and ideas into a project by topic. In the new chat, ask what's open on my project? and Claude reads them back. Saving again later adds only what's new. Workbranch has a 3-day free trial at app.workbranch.ai.

Sources

Frequently asked questions

Is the length limit the same as the usage limit?

No. The length limit caps how much one conversation can hold. The usage limit caps how much you can do across all chats in five-hour and weekly windows. You can hit either one without the other.

Will Claude forget the old chat if I start a new one?

The new chat starts with its own context. On paid plans with chat search switched on, Claude can look up the earlier conversation when you ask, and inside a project it searches that project's chats. Otherwise, bring the important parts over yourself with a summary.

Does a bigger model help?

It can. Anthropic says the context window depends on the model; on paid plans the newest models go up to 1M tokens while others have 500K or 200K. A larger window only delays the limit, though.