AI Helper — copy a conversation out of the extension
Sometimes a conversation with the AI Helper on one domain reveals a solution worth moving elsewhere — to another tab, to Claude.ai for a follow-up, to your team docs, to a tracker issue. In v3.2.81 we added a dropdown to the AI Helper window header with two copy options: the whole conversation (Markdown) or the last API input (with the full conversation context). One click, context preserved.
Why a dropdown
Up to v3.2.80 the AI Helper had a single "copy" button in the header — it copied the last model message. That was fine if you wanted to move one sentence of the answer — but it lost the entire conversation context. Users kept asking to copy the whole conversation as a record of what they did — questions and answers in one note. Hence the expansion into a dropdown with two levels of detail.
Option 1 — Copy the whole conversation (Markdown)
Exports every turn of the conversation as clean Markdown:
**You:**
Hide the right-hand suggestions on YouTube watch pages.
**AI Helper:**
I'll target #related and add `display: none`.
…proposed code…
**You:**
Apply it.
**AI Helper:**
Done — rule saved as "YouTube: hide suggestions" in folder "Personal".The format is ready to paste into Notion, Linear, a GitHub issue, any messaging app or a text document. Code blocks inside are real Markdown (fenced with ```) — formatting doesn't break.
Option 2 — Copy the last API input (with conversation context)
The second option exports the full payload of the last API call — system prompt, conversation history, any tool calls, the most recent DOM snapshot. Paste it into another AI window (Claude.ai web app, OpenAI playground) and you can keep the same task with a different model, or negotiate a different solution without losing context.
Sample use: the AI Helper got stuck in a loop of suggestions that don't work. You copy "last API input" into Claude.ai, add the prompt "try a completely different approach" — Claude (with a different model, different tuning) sometimes proposes something the in-extension model didn't find.
Three use cases where this matters
1. Backup of a successful workflow. You finished a conversation that solved a hard problem (CSS for an aggressive page, JS for a SPA). Click "whole conversation", paste it into your notes — next time you hit a similar problem, you have a working template.
2. Move an idea between domains. You cracked something on facebook.com — you want to repeat it on x.com. Copy the conversation, open the AI Helper on x.com, paste "I want to do the same thing as here" — the context lives outside the first domain.
3. Second opinion from a different model. Your AI Helper workspace uses Claude Sonnet, but you want to try GPT-4o or Gemini Pro on the same task. Copy the "last API input", paste it into their playground — compare answers side by side.
What the export contains and what it doesn't
Exported:
- Every user and model message body.
- Markdown formatting for code blocks and lists.
- The system prompt (with "last API input") — the AI Helper's role definition and available tools.
- The most recent page DOM snapshot (with "last API input").
Not exported:
- Your API keys (they live in the extension's local storage, not in the conversation body).
- Other model settings (max_tokens, temperature, etc. — they sit in the request body, not in the content).
- Your personal data or history from other conversations.
The export is exactly what you already sent to the AI provider in your last turn — no extra data leaks out of the extension.
See also
- On-page windows — the AI Helper as one of the developer windows
- Tokens with cache in the AI Helper — real cost numbers for a conversation
- AI Helper in a shadow DOM — the recent style isolation from the page
Install JustZix — and move conclusions from AI Helper conversations in whatever context fits you.
Rate this post
No ratings yet — be the first.