ChatGPTGeminiClaudeGrok40,000+users on the Chrome Web Store4.7/ 5

ChatGPT for Data Scientists, Filed by Experiment

Data scientists use ChatGPT for pandas and Polars idioms, scikit-learn pipelines, statistical test selection and PyTorch error traces, across several experiments at once. The answers land in one flat sidebar with no experiment context and no way to find the loss function from the ranking project. AI Toolbox (formerly ChatGPT Toolbox), a Chromium extension by Infi Developments, adds a folder per project and experiment, full-text search across every message, a // prompt library with chains for repeatable analysis, and Markdown export whose frontmatter records the model that produced the answer.

Free with 2 root folders, 5 search results per query and 3 exports a day. Premium removes the caps and adds the Advanced search filters, bulk export and prompt chaining.

Add to Chrome, freeSee pricing

Free plan, no account | Premium $9.99/mo | Lifetime $99 | All Access $199 once for all four modules

Manage Folders panel in ChatGPT listing Data Analysis, Methodology and Literature Review folders with chat, subfolder and GPT counts

Manage Folders on chatgpt.com

  • A folder per workstream (Data Analysis, Methodology) with subfolders per experiment or dataset underneath
  • Pin the experiment you are iterating on so it stays at the top of the sidebar
  • Search folders by name; each row shows its chat, subfolder and GPT counts

The gap

What ChatGPT gives a data scientist, and what it loses between experiments

ChatGPT has Projects, a title-only search and one all-account export. None of them keep the churn model apart from the recommender, find the cross-validation discussion from March, or record which model wrote the answer you are about to reproduce.

Modeling work is repetitive in structure and unique in data: the same EDA questions, test-selection logic and error-trace prompts, a different dataset each time. In ChatGPT the good prompt is buried in an old chat, the custom loss derivation is a reply you cannot search for, and the experiment boundary is whatever you remember. Projects hold chats one level deep and cannot search inside them.

AI Toolbox keeps the ChatGPT interface and adds the missing layer in your browser: a folder tree per project and experiment, an index of every message you can query by dataset, metric or method name, saved prompts and chains you insert by typing //, and per-conversation export as Markdown with title, date, model, URL and source in the frontmatter. Nothing is uploaded; the index stays on your machine.

Why it matters

Here is what you cannot do with ChatGPT alone:

  • Keep experiments apart. Projects are flat, unpinned and cannot nest EDA, feature engineering, modeling and write-up under one project.
  • Find last quarter's method. Native search matches chat titles, not the function names, metrics or test names inside replies.
  • Rerun an analysis the same way. There is no prompt library or chain; the EDA sequence is retyped from memory on every new dataset.
  • Record which model answered. The native export is your entire account as HTML in a ZIP, with no per-conversation file and no model field to filter on.

The longer a project runs, the more of the methodology sits in old replies. Folders, search and a prompt library are what make it reproducible on the next dataset.

Step by step

A data science workflow in five steps

From install to a notebook-ready export. Steps one to four are on the free plan; chains, Markdown and bulk export in step five are where Premium starts.

About three minutes

  1. Install AI Toolbox from the Chrome Web StoreWorks in Chrome, Edge, Brave, Opera and Arc. Open chatgpt.com and the folder tree, search and prompt controls appear in the sidebar.
  2. Create a folder per project, subfolders per experimentManage Folders, then Add Folder: Churn Model, with EDA, Features, Modeling and Write-up underneath. Pin the experiment you are iterating on.
  3. File each conversation as you goOpen the project folder, search for the conversation by name and add it. Bookmark the reply that holds the encoding decision or the hyperparameter rationale you will cite.
  4. Save the prompts that worked, then chain themType // to insert a saved test-selection or error-trace prompt. Turn your EDA sequence into a chain with a {{dataset}} variable so it runs the same way on the next table.
  5. Search, then export for the notebookCmd/Ctrl+Shift+F and type the metric or function name. Export the conversation as Markdown for a notebook cell or Obsidian page; the frontmatter records the model. Bulk export the whole project folder at write-up.
Add to Chrome, free

Advanced search on chatgpt.com

Advanced search panel with exact match, role and date filters open above the results

Exact match, role and date filters above the results. Search for a function name with exact match on, limit it to assistant replies, and narrow the date range to the sprint you ran the experiment in. Advanced filters are a Premium feature; plain full-text search with five results per query is on the free plan.

The four features data scientists use

Experiment folders, search, prompts and export, as they ship

Every capture below is the extension running inside chatgpt.com. The plan each feature needs is stated in the text.

Folders

A folder per project, subfolders per experiment

Nested subfolders in the ChatGPT sidebar, added by searching a conversation's name rather than dragging. Mirror your repo: a project at the top, EDA, features, modeling and write-up below it. Free covers two root folders with unlimited subfolders; Premium removes the cap and adds bulk move.

AI Toolbox folder tree with nested subfolders inside the ChatGPT sidebar

Search

Find the derivation by function, metric or test name

The index covers every message in your history, so a loss function name, a metric or a statistical test finds the reply that contains it, highlighted, with a link into the chat. Cmd/Ctrl+Shift+F opens it anywhere on chatgpt.com. Free returns five results per query; Premium adds unlimited results with exact match, role and date filters.

Search History panel showing a query matched in 11 messages across 3 conversations, with per-message highlights and bulk actions

Prompts

Type // to load a modeling prompt

Keep your EDA, test-selection, pipeline-review and error-trace prompts in the library and insert one by typing // in the compose box, with {{dataset}} or {{target}} placeholders filled at insert time. The free plan stores two prompts; Premium makes the library unlimited and adds prompt chaining for multi-step analyses.

Inserting a saved prompt by typing // in the ChatGPT compose box

Export

Markdown with the model in the frontmatter

Any conversation exports as Markdown, PDF, JSON or TXT from a dialog on the chat itself. The Markdown file opens with YAML frontmatter (title, date, model, URL, source), so it pastes into a Jupyter Markdown cell, Obsidian or Logseq and can be filtered by model later. Free exports as TXT, three a day; Premium unlocks Markdown, JSON and PDF, removes the cap and adds bulk export of a whole project folder.

Select Export Format dialog with Text, Markdown, JSON and PDF options

Scenarios

Where it pays off

Three data science workflows and what to file, save and export in each.

  • EDA and feature engineering

    • Project folder, subfolder per dataset
    • Bookmark the reply with the encoding decision
    • Save the pandas and Polars idioms you keep re-asking
    • Export Markdown into the notebook

  • Model debugging and tuning

    • Subfolder per model family
    • Search the error message before asking again
    • Chain describe, diagnose and fix steps
    • Pin the run you are iterating on

  • Statistical analysis and write-up

    • Save the test-selection prompt with a {{target}} variable
    • Search by test name across experiments
    • Export Markdown with the model recorded
    • Bulk export the project folder for the methods section

  • Several experiments, one browser

    Everything stays on your machine; nothing is uploaded. See how the ChatGPT folders work.

    Add to Chrome, free

From the Chrome Web Store

What reviewers say about search and export

Public reviews from the listing, shown as captured. The full wall is on the reviews page.

  • Chrome Web Store review by Asaf Ben-Yehuda: makes chatgpt search and usage much better. helps you find what you want to find. glad i joined when this add on was first introduced!
    Asaf Ben-Yehuda
  • Chrome Web Store review by Dave G: Great extension! Love the export feature.. The support is good, very responsive and helpful!
    Dave G

Read all 46 captured reviews or the live Chrome Web Store listing.

Pricing

Plans and pricing for the ChatGPT module

Free covers folders, search, prompts and export with usage caps. Premium removes the caps on ChatGPT and adds Advanced search, bulk export and prompt chaining. All Access does the same across ChatGPT, Gemini, Claude and Grok, paid once. Not an AI subscription: you keep using your own ChatGPT account.

Free

$0forever

  • 2 root folders with unlimited subfolders, 2 pinned chats
  • Full-text search, 5 results per query
  • 2 saved prompts with the // shortcut
  • 3 TXT exports a day
Add to Chrome

Premium

$9.99/month

  • Unlimited folders, pins, prompts and exports in Markdown, JSON, PDF or TXT
  • Advanced search: exact match, role and date filters, unlimited results
  • Bulk export and bulk move to folder
  • Prompt chaining and cross-platform search
Get Premium

LifetimeMost popular

$99one-time

  • Everything in Premium
  • Pay once, use forever
  • All future updates
  • Top priority support

Pays back in under 10 months of Premium.

Get Lifetime Access

All Access LifetimeBest value

$199one-time

  • ChatGPT, Gemini, Claude and Grok included
  • Search all platforms across all four
  • Every future AI platform we add
  • Save $197 vs single Lifetimes
Get All Access

Teams$15/seat/month

All Premium features across the four modules · Admin dashboard · Team analytics · $12/seat/mo billed annually · 14-day free trial

See Teams
  • GDPR compliant
  • Secure checkout by Polar
  • 14-day money-back on a first purchase
  • Conversation content stays in your browser

Refunds follow the refund policy; renewals are not refundable. Premium unlocks on the ChatGPT account whose email you enter at checkout, so use the address you sign in with.

FAQ

Questions people ask before installing

How do data scientists organize ChatGPT conversations by experiment?

Create a root folder per project and subfolders per experiment, model family or dataset (EDA, features, modeling, write-up), then add conversations by searching their name inside the folder. Pinned folders keep the run you are iterating on at the top of the sidebar. The free plan includes two root folders with unlimited subfolders; Premium removes the cap and adds bulk move.

Can I export ChatGPT conversations into Jupyter or Obsidian?

Yes. Any conversation exports as Markdown with YAML frontmatter (title, date, model, URL, source), which pastes into a Jupyter Markdown cell, an Obsidian vault or a Logseq page without reformatting. PDF, JSON and TXT are also available. The free plan exports as TXT, three a day; Markdown, JSON and PDF need Premium, which is unlimited and can export a whole project folder at once.

Does the export record which model produced the answer?

Yes. The Markdown frontmatter carries a model field with the model slug ChatGPT reports for the conversation, alongside the title, date, URL and source. That lets you filter an archive by model later or check which model produced a result before you reproduce it.

Can I find a method I worked through months ago?

Yes. AI Toolbox indexes every message in your history, so a function name, a metric or a statistical test returns each reply that contains it, highlighted, with a link into the chat. Press Cmd/Ctrl+Shift+F on chatgpt.com. Free shows five results per query; Premium shows all of them with exact match, role and date filters.

Can I chain prompts into a structured analysis?

On Premium, yes. Save the prompts as a chain with variables such as {{dataset}} or {{target}}, and the sequence (describe the data, propose tests, recommend a model, draft the code) runs in the same order every time with only the specifics changed. Single saved prompts and the // shortcut are on the free plan.

Is AI Toolbox safe to use with proprietary data?

The folder tree, search index, bookmarks and saved prompts are stored in your browser, and every search runs locally. Conversation content is never sent to our servers; the only account detail that reaches them is the email used to verify a Premium plan. What you paste into ChatGPT is still governed by OpenAI's terms and your data agreements, as it would be without the extension.

Which browsers work?

Any Chromium browser installed from the Chrome Web Store: Chrome, Edge, Brave, Opera and Arc. There is no Firefox version.

Related guidesHow to organize ChatGPT with foldersSave reusable ChatGPT promptsAutomate ChatGPT prompt sequencesExport ChatGPT conversationsChatGPT for developersWhy data scientists use AI Toolbox

40,000+ users · 4.7/5 on the Chrome Web Store

Start the next experiment with the last one's methodology one search away

Install AI Toolbox, make one folder for the model you are on, and save the EDA prompt you just rewrote for the third time before you open the notebook.

Add to Chrome, freeSee the ChatGPT module