How to Move a Claude Conversation to a New Chat Without Losing Context

Why a Fresh Chat Beats Pushing On

Long conversations degrade in predictable ways: early constraints get diluted, superseded drafts confuse which version is current, and very long threads may be summarized automatically, which loses detail. Anthropic's own guidance for long chats approaching a limit is to start a new conversation. The catch is that a new conversation starts with nothing, so the quality of the move depends entirely on what you bring with you.

A good handoff is not a transcript dump. Pasting 40,000 words of old chat into a new one recreates the exact problem you were escaping, and it costs usage on every following message. The goal is to compress the thread into the smallest set of facts that lets Claude do the next piece of work as if it had been there the whole time. If you are unsure whether to move at all, see why Claude loses track in long chats.

What to Carry Over and What to Leave Behind

The "rejected options" row is the one most people skip, and it is the one that saves the most frustration. Without it, the new chat will often suggest the first idea you threw out two hours ago.

A Copy-Paste Handoff Brief Template

Use this as the first message in the new chat. Fill in each section with short lines. Most good briefs land between 200 and 600 words plus the current artifact.

I'm continuing work from a previous conversation. Treat this brief
as the full context. Do not assume anything not written here.

## Goal
[One or two sentences: what we are producing and for whom.]

## Decisions made
- [Decision] — because [reason]
- [Decision] — because [reason]

## Rejected (do not re-propose)
- [Option] — rejected because [reason]

## Constraints
- [Format / tone / length / stack / audience]

## Current state
[Paste the latest version of the draft, code, or table here.]

## Open questions
- [Question]

## Next step
[The specific thing to do now.]

Before doing the next step, summarize your understanding in
5 bullets and list anything that seems ambiguous. Then wait.

The closing instruction matters. Asking Claude to restate the brief and flag ambiguities costs one short turn and catches most misreadings before they turn into a wrong draft.

Step by Step: Moving a Thread

1. Pick a clean break point. Move after a milestone (outline approved, function working), not in the middle of a half-finished change.
2. Ask the old chat for a draft brief. Claude can summarize its own thread well while the context is still coherent. Use the template above as the requested format.
3. Edit the draft yourself. Check every decision against your memory of the thread. Remove anything that is not needed for the next step. Add rejected options if they are missing.
4. Paste the latest artifact verbatim. Do not ask Claude to "remember" the final draft. Copy the exact text in.
5. Start the new chat in the right place. If the work belongs to a project, open the new chat inside that project so project instructions and knowledge load automatically.
6. Confirm, then continue. Read Claude's restatement. Correct anything off before asking for real work.
7. Keep the old chat. Do not delete it. If you need an exact quote or a detail you dropped, you can go back and find it.

Worked Example: A Pricing Page Thread

Suppose you have spent a long session drafting a SaaS pricing page. The thread contains three tier structures, four rounds of copy, a comparison table that changed twice, and a long tangent about annual discounts. Claude has started mixing tier names from the first and third structures.

The brief you carry over looks like this:

## Goal
Pricing page copy for a B2B analytics tool. Audience: ops leads
at 10-50 person companies.

## Decisions made
- Three tiers: Starter, Growth, Scale — Growth is the anchor
- Monthly pricing only at launch — simpler to test
- Feature table: 8 rows max — longer tables tested poorly with users

## Rejected (do not re-propose)
- Free tier — support cost too high for current team size
- Annual discount banner — revisit after launch

## Constraints
- Plain language, no "unlock" or "supercharge"
- Each tier description under 25 words

## Current state
[Final v4 copy pasted here]

## Next step
Write the FAQ section: 6 questions, objection-focused.

That is roughly 120 words of brief plus the copy. The new chat starts with every decision that matters, none of the noise, and an explicit next task. In practice that usually produces a better FAQ on the first try than asking the original, overloaded thread would have.

Adjusting the Brief for Code, Research, and Writing

The template above is general. Different kinds of work need different emphasis in the "current state" and "constraints" sections, and the wrong emphasis is the most common reason a brief underperforms.

For code, be especially careful with the current state. A brief that describes the code ("the auth middleware checks the token") is much weaker than one that pastes it. Claude can only match conventions it can see. For research, carry over the sources themselves, not Claude's paraphrase of them, so the new chat can verify claims instead of repeating them.

If you regularly hand off the same kind of work, save your adjusted template as a reusable prompt. Writing the brief then becomes a fill-in-the-blanks job that takes a few minutes.

Common Mistakes When Moving a Chat

• Trusting the summary blindly. Claude's self-summary of a drifted thread can inherit the drift. Always review it.
• Carrying over every version. Only the current version goes in. If an older version matters, say why in one line.
• Leaving out reasons. "No free tier" without "because support cost" invites Claude to argue for one again.
• Making the brief the new transcript. If the brief is longer than the work, it is too long.
• Forgetting reusable context belongs elsewhere. Style guides and reference docs you paste into every brief should live in project knowledge or account-level instructions instead.

When not to move: if you only need one or two more replies, a quick restatement in the existing thread is faster. And if you want to explore an alternative without losing the current thread, forking is a better tool than starting over. See fork vs branch in Claude.

ClaudeKit Workflow Note

ClaudeKit's Handoff Brief tab, inside the Workspace panel on claude.ai, turns a long conversation into a compact brief you can paste into a new chat. Free users get two briefs per day. It pairs well with pinning: pin the messages that hold real decisions as you go, and you have a quick way to check the brief against what was actually agreed.

If you are not ready to retire the thread, ClaudeKit can also fork from any message. Forks are free and unlimited, and the original conversation stays intact. Treat the brief as a starting point either way: read it, cut it, and add the rejected options before you paste.