Clio

Clio is a Chrome extension that extracts your AI conversations to structured JSON — text, code, thinking, images — and saves them to your local disk. The data does not leave your browser. There is no telemetry. The source is on GitHub.

Install

Coming soon to the Chrome Web Store.

In the meantime, Clio can be loaded from source — see the GitHub repository for instructions.

Three commitments

Clio's design is organized around the security CIA triad. Each commitment is enforced by the manifest, not by a runtime promise.

• Confidentiality — Your conversations never leave your browser. No remote transmission, no telemetry, no third parties. Privacy policy.
• Integrity — What Clio captures faithfully reflects what you saw, or extraction fails loudly. Silent partial captures are unacceptable. Security policy.
• Availability — Your access to your own conversations does not depend on the provider's continued goodwill. Why this matters.

How it works

1. Open any Gemini, Claude, or ChatGPT conversation in your browser.
2. Click the Clio toolbar icon.
3. Click Extract Conversation.
4. Save the ZIP — a structured conversation.json plus an images/ folder.

What Clio captures

• Every message in the conversation, in DOM order, with roles preserved
• Code blocks with language labels
• Embedded images, fetched at extraction time
• Assistant thinking / reasoning sections when expanded
• Conversation metadata — title, source URL, timestamp

Permissions

Clio's manifest requests the minimum surface area required to do its job:

• activeTab — read the DOM of the tab you're currently looking at, only when you press the toolbar button
• downloads — write the extracted ZIP to your local disk via the "Save As" dialog
• Host access to gemini.google.com, claude.ai, and chatgpt.com — and nothing else

No tabs, no cookies, no webRequest, no storage, no <all_urls>. The manifest is the ground-truth permission declaration.

Open source

Clio is published under the MIT License. The full source code is at github.com/martymcenroe/Clio.

For governance and design discussion — why these design choices, what threat model they address, what they don't address — see the project wiki.

Contact

Privacy questions or feedback: [email protected].

Security vulnerabilities: see SECURITY.md for the private reporting channel. Do not file public issues for security reports.