◉CLAUDE LABJP
●2.1.302 — Fork subagents are now on by default in claude -p, the Agent SDK and the desktop app. Turn them off with CLAUDE_CODE_FORK_SUBAGENT=0 (10/16)●2.1.307 — Fixes MCP failing to reconnect after sleep or a keychain lock (10/21). 2.1.306 also fixed the false claude -p login message●SONNET 4.5 — Retires on the Claude API on 11/30, 38 days left. Migrate to Sonnet 5.5●WINDOWS — Reports say hooks break silently only on Japanese-locale Windows; check the character encoding first●FAST MODE — Opus 5.5 fast mode is out of research preview and no longer needs the beta header (10/15)●NEW — I Put a Ten-Line Promise File at the To…●2.1.302 — Fork subagents are now on by default in claude -p, the Agent SDK and the desktop app. Turn them off with CLAUDE_CODE_FORK_SUBAGENT=0 (10/16)●2.1.307 — Fixes MCP failing to reconnect after sleep or a keychain lock (10/21). 2.1.306 also fixed the false claude -p login message●SONNET 4.5 — Retires on the Claude API on 11/30, 38 days left. Migrate to Sonnet 5.5●WINDOWS — Reports say hooks break silently only on Japanese-locale Windows; check the character encoding first●FAST MODE — Opus 5.5 fast mode is out of research preview and no longer needs the beta header (10/15)●NEW — I Put a Ten-Line Promise File at the To…
Articles/Claude.ai
◉ Claude.ai/2026-10-23Beginner

Moving a Long Claude Chat to a New One: The Three Items I Keep in the Handover Note, and What I Leave Out

When a long Claude conversation gets slow or eats your usage, a full summary is the wrong thing to carry over. Keep decisions, rejected ideas and the next step, and use this ready-made note template.

claude-ai17long conversationscontext12handoverusageworkflow42

I once spent several days polishing an app's store description inside a single conversation. At first it went well, and Claude picked up my phrasing habits nicely.

Then, somewhere around day three, replies slowed down and my remaining usage started draining faster than I expected. I knew that starting a fresh chat would make things lighter. But I was afraid to leave behind everything we had built up, so I dragged the same conversation along for longer than I should have.

Here is what I want to tell you first: moving to a new chat is easy; deciding what to carry over is the hard part. As an indie developer, I got this wrong a few times before settling on a method that keeps just three items. This post is a record of that method.

"Heavy" can mean two different things

Before thinking about handovers, it helps to sort out the symptom. Claude's Help Center describes two kinds of limits.

TypeWhat it controlsTypical symptom
Usage limitsHow much you can use Claude over a period (a session limit that resets every five hours, plus weekly limits)A "limit reached" message, and a wait until reset
Length limitsHow long a single conversation can grow (the context window)One conversation becomes hard to continue, or older content gets harder to work with

The quick test: if every conversation stops, it is a usage problem; if only one conversation is sluggish, it is a length problem. This article is about the second case, though the first tends to tag along.

According to the official explanation, longer conversations consume more of your usage. If code execution is enabled, Claude also summarizes earlier messages automatically as the window fills up. That is handy, but the documentation notes that such long chats cost more usage, and suggests starting a new conversation when you are close to your limit.

So moving on once a chat feels heavy is not just a hunch. It matches the official guidance.

What I tried first, and why it failed

My first approach was to end the old chat with "summarize everything so far," then paste the result into a new one.

It did not go well. The summary was careful, and also long. Once pasted, it took up most of the new conversation's opening, and the one thing that mattered, what to do next, got buried.

The second failure was subtler: the summary did not mention the ideas I had rejected. The new Claude happily proposed phrasings I had already thrown away, and I ended up repeating the same exchange from the start.

The rule I drew from that fits in one sentence:

Choose what goes into the summary not by "what happened," but by "what the next conversation needs so it does not trip in the same place."

The three items I keep

My handover note now contains only these three things.

1. Decisions (things I will not reopen)

Tone, target reader, length cap, terms to use and terms to avoid. I write them as short bullets, conclusions only, with no backstory.

2. Rejected ideas, with the reason

This turned out to be the most effective item. One line each, such as "too stiff, dropped." With the reason attached, Claude also steers clear of similar ideas.

3. The next step (the first request in the new chat)

Exactly one thing. If this is vague, Claude starts by commenting on the note itself.

The template I actually use

At the end of the old conversation, I ask:

I'm moving this conversation to a new chat.
Summarize only the following three items, within 400 characters total.
Skip the backstory and any recap of what went well.
 
1. Decisions (things that will not change)
2. Rejected ideas, with a one-line reason each
3. The next step (one request to make at the start of the new chat)

The problem it solves is the summary that grows so long it takes over the new chat's opening. I cap the length up front because otherwise Claude, being helpful, writes more. Keeping it short is what makes the handover good.

I paste the result as the first message of the new chat. Before pasting, read it once yourself and check that the decisions are accurate.

What I deliberately leave out

  • The flow of the conversation: who said what in which order does not help a new chat.
  • Intermediate drafts: only the adopted version belongs under decisions. If you need a draft, save it as a separate file.
  • Claude's long explanations: if you need them, you can ask again.

When I hesitate, I ask myself: "Without this, would the next conversation repeat the same mistake?" If yes, keep it. If no, drop it.

Avoiding the long chat in the first place

Even with a good handover, creating long conversations repeatedly is a chore. Two measures from the Help Center helped me most.

First, put reference material you reuse into a Project. Projects load only the relevant parts, so they keep conversations lighter than pasting the same material every time. I cover this in Claude Projects Practical Guide.

Second, turn off extended thinking and connectors you are not using. The documentation notes that tools and connectors are token-intensive, so switching them off from the start for conversations that need no research makes a visible difference.

If you want to plan your day around usage, see planning a day around the five-hour session window. And if Claude seems to lose instructions mid-conversation, separating forgetting from not seeing covers a related problem.

Next time a chat feels heavy

I first sort out the symptom, and if only one conversation is heavy, I move. At the end of the old chat I ask for the three items within 400 characters, then paste them as the first message of the new one.

You only need to do one thing: open a conversation that feels heavy right now and write three lines, "decisions, rejected ideas, next step." If you can write them, that is already your handover note.

You may be surprised how much of the old chat you can let go of.

Share

Thank You for Reading

Claude Lab is ad-free, supported entirely by members like you. We publish practical guides daily with implementation code, benchmarks, and production-ready patterns. If you've found it useful, we'd love to have you on board.

  • ✦Copy-paste ready implementation code
  • ✦New advanced guides published daily
  • ✦$5/mo or $10 for lifetime access
View Membership →

If you found this article helpful, a small tip ($1.50) would mean a lot to us. Your support helps keep this site ad-free and covers server and hosting costs.

Related Articles

◉ Claude.ai2026-10-15
Let Claude Draft the Quote's Wording, Keep the Numbers in Your Own File — A Small Source of Truth and a Check Script
How a freelancer can have Claude polish the wording of quotes and invoices while the amounts stay in a JSON file and get verified by a short Python script.
◉ Claude.ai2026-09-17
What I Stop Doing Before I Hit Claude's Limit — Rebuilding a Day Around the Five-Hour Window
Claude's usage recovers on a five-hour window, but that window doesn't start on the clock — it starts with the first message you send. Here's how I reordered my working day so I stop before the limit stops me.
◉ Claude.ai2026-09-15
Where to Put an Instruction That Keeps Coming Back: Chat, Style, Settings File, or Hook
No comments, please. Do not translate the product name. When an instruction drifts back a few turns later, the fix is usually the location, not the wording. I compared the same sentence across four layers and settled on a rule for which instruction belongs where.