commentable<html>
Works with Claude · ChatGPT · Gemini CLI

Your AI writes it.
commentable ships it.

Add commentable to Claude, ChatGPT, or Gemini CLI as a connector. Then just ask — your AI publishes the HTML it wrote as a private, encrypted link your team can comment on. No copy-paste, no leaving the chat.

claude.ai
+ New chat
Here is your dashboard:
Q3 Revenue · Kestrel

Want me to publish it as a commentable link for your team?

Yes. Expire in 7 days, no password.
✓commentable · create_commentable_link
Published with your branding:share.acme.studio/a/k3mX9pQr#key=aG9wZQ
# design-review Dana, Sam, Mia + 9
Y

You 2:41 PM

Q3 dashboard is live — comments welcome 👇

Q3 Revenue · Kestrel

share.acme.studio · end-to-end encrypted

Open the live page →

share.acme.studio/a/k3mX9pQr
Acme Studio💬 0
Q3 Revenue · Kestrel v2
DDana · Q3 target line here?
SSam · love this trend 🔥
MMia · swap axis labels
v2 published — the link never changed
Feedback in minutes. On the work itself.
Same link. Every version.

commentable <html>

commentable <html>

Share HTML artifacts as live, commentable links.

html.commentable.link

The whole flow: build in Claude, answer expiry and password, get your branded link, and watch the comments land. Claude always asks before publishing — you stay in control.

Setup

Two minutes, three steps.

  1. 01

    Create your API key

    Sign in and open your dashboard → AI connectors → pick your assistant. We create the key and open the right dialog for you. One key works everywhere, links published pages to your account, and can be revoked anytime. The steps below show what that dialog looks like.

  2. 02

    Connect your AI

    On claude.ai (Pro, Max, Team, Enterprise)

    SettingsConnectorsAdd custom connector
    Add custom connector✕
    commentable
    your connector URL
    https://html.commentable.link/api/mcp?key=YOUR_API_KEY

    After connecting, set the Publish HTML as a commentable link tool to Always allowso publishing doesn't pause each time, then start a new conversation.

    In Claude Code

    One command, and the key rides an Authorization header (out of URLs and shell history):

    claude mcp add --transport http commentable https://html.commentable.link/api/mcp --header "Authorization: Bearer YOUR_API_KEY"

    Ready to connect? Your dashboard creates the key and opens the right dialog for you, one click.

  3. 03

    Just ask

    “Create a commentable link for this page.” Your AI asks about expiry and an optional password, publishes, and drops the live link into the chat. When feedback lands, say “read the comments on my link” and it pulls every pinned comment in — then fixes the page and updates the same URL. The page appears in your dashboard like any other.

Live data

A dashboard that updates itself.

Claude artifacts can pull live data through connectors — but those calls only run inside claude.ai, and connector artifacts can't be shared publicly at all. The connector handles this for you: when you publish, Claude bakes the current data into the page as a static snapshot. Add a scheduled task and the same link serves a fresh snapshot every day.

Set it up once

A scheduled task starts a brand-new conversation, so its prompt has to carry everything: what data to fetch, how to rebuild the page, and which link to update. When you publish a dashboard that uses connectors, Claude asks whether you want auto-refresh and hands you this prompt ready-made, your link already inside. If you are setting it up by hand, this is the shape:

Refresh my [dashboard name] page: 1. Fetch the latest [metrics] using the [connector name] connector. 2. Rebuild the complete dashboard HTML with the new data inlined — a static page, no window.claude calls. 3. Call update_commentable_link with url = [paste the FULL link including #key=] and the new HTML. Update in place — do not create a new link.

Save it as a scheduled task in Claude or ChatGPT (daily, weekly, whatever the data deserves) with the commentable connector enabled. The link you shared never changes, pinned comments survive every refresh, and viewers always see the latest numbers.

The #key= fragment is the page's decryption key — paste the full link only into your own scheduled task, never anywhere public.

Encrypted, with keys you control.

Connector publishes are encrypted on our server, in memory: each one generates a fresh AES-256-GCM key that lives only for the duration of the request — it's embedded in the link's #fragment and never stored or logged. Our database only ever holds ciphertext. (Pages you create at html.commentable.link are encrypted in your browser instead — the strongest mode.)

Your API key authorizes publishing under your account — it can't read existing pages, and you can revoke it in the dashboard at any time.

Treat the connector URL like a password: it contains your key.

Which AI assistants work with commentable?

Claude (claude.ai on Pro, Max, Team, Enterprise — and Claude Code), ChatGPT (Plus and Pro via Developer Mode; Business and Enterprise when an admin enables it), and Gemini CLI. One API key works for all of them.

How do I connect it to ChatGPT?

Turn on Developer Mode under Settings → Apps & Connectors → Advanced settings, then create a custom connector with your commentable URL and No Auth. Two minutes, and it works in scheduled tasks too.

Can I update a published page later?

Yes — ask your AI to update the link, or open the page and hit Update HTML. The link (and everyone who has it) stays the same.

Does the AI decide the expiry or password?

No. The connector instructs every assistant to ask you both questions before publishing.

What if my API key leaks?

Revoke it in the dashboard — every connector using it stops working instantly. Already-published pages are unaffected.

From prompt to feedback in one message.

Get your API key