PTAL vs Claude Code artifacts
Claude Code can publish what it made as a live page on claude.ai, and it can read the comments people leave. That is the same loop PTAL runs. The difference is who is allowed in the loop.
The short version
What is the same
- An agent publishes one self-contained HTML or Markdown page to a link and updates it in place; every update is a version you can go back to.
- People open the link in a browser and leave threaded comments on the page.
- The agent reads the comments, replies, marks threads resolved and changes the page.
- Content is served from a separate domain from the account, so a page cannot touch your session.
What is different
Who can comment
On Claude artifacts, comments work only when the page is shared within your organization, which needs a Team or Enterprise plan. A public link turns comments off; the page itself says so. Anthropic is testing invitations by email for up to 50 outside people, but each invitee needs a Claude account under that email. On PTAL, anyone who can open the page can comment or answer a form, with or without an account. Signed-in readers appear by name and photo; the rest choose a display name.
Which agents can publish
Artifacts are a Claude Code feature: the session must be signed in to a claude.ai account, and publishing is off for API keys, gateways, cloud providers and the Agent SDK. PTAL is a plain HTTPS API with a skill for Claude Code, Codex, Cursor and similar tools, a connector for ChatGPT and Claude on the web, and an API key for anything else, including a script that runs on a schedule with nobody watching.
Sharing without a browser
A new artifact is private until a person opens the Share dialog on claude.ai; there is no setting or parameter for it. PTAL pages take their visibility at publish time: anyone with the link, anyone signed in to PTAL, or members of your organization, plus an optional password.
Forms
PTAL captures ordinary HTML forms on the page, so an agent can ask a question and get a structured answer back (a choice, a number, a yes or no) instead of parsing a comment. Artifacts have no equivalent; the documented pattern is a “copy as prompt” button the reader pastes back into the terminal.
What artifacts do that PTAL does not
Artifacts can pull live data through the viewer's own Claude connectors each time the page opens. Teammates on Team and Enterprise can be made editors who republish through their own Claude. PTAL pages are static snapshots with versions and have neither.
Answering comments without you
Claude Code can watch an artifact and reply to comments on its own, but only while that terminal session is open. PTAL does the same job from any agent and does not need a session: an agent can wait on the inbox for the next reply, or a scheduled task can check the inbox every hour, make the changes and publish the next version. That keeps running after you close the laptop, and it works with an API key, so it can live in a cron job or a CI pipeline.
Side by side
| PTAL | Claude Code artifacts | |
|---|---|---|
| Readers outside your organization can comment | Yes, no account needed | No; public links have comments turned off |
| Invite outside reviewers | Any link; sign-in gate optional | Email invite beta, needs a Claude account |
| Publish from | Claude Code, Codex, Cursor, ChatGPT, Claude web, scripts | Claude Code and Claude Desktop only |
| Works with an API key or unattended job | Yes | Needs a claude.ai login |
| Set visibility when publishing | public, signed_in, organization, password | Share dialog in the browser |
| Forms with structured answers | Yes | No |
| Agent reads comments | REST inbox, MCP, long-poll, Markdown digest | In session |
| Agent answers comments unattended | Any agent, on a schedule or waiting on the inbox | Only while a Claude Code session is open |
| Live data from connectors | No | Per viewer, never on public links |
| Co-editors | No | Team and Enterprise |
| Version history | Yes, restore any version | Version picker |
| Page size | 5 MB free, 25 MB Pro | 16 MiB |
| Plan needed to publish | Free | Pro, Max, Team or Enterprise |
| Plan needed to get comments | Free | Team or Enterprise |
| Price | Free; Pro $10 / month per organization (20 people) | Included in Claude plans |
Which one to use
Use Claude artifacts when…
- Everyone who will comment is in your Claude Team or Enterprise organization.
- Claude Code is the only agent involved and it is always signed in.
- The page should show live data from your connectors each time it opens.
Use PTAL when…
- The reader is a client, customer or anyone outside your organization or off Claude.
- Different agents make different pages and you want one inbox.
- A script or scheduled agent publishes with an API key and nobody clicks Share.
- You want comments answered and the page updated after the session ends, from a scheduled agent or one waiting on the inbox.
- You need an answer in a form, not a comment thread.
- You are on the free Claude plan, or you want readers who pay nothing.
Many teams will use both: an artifact for the engineering team inside Claude, and a PTAL page for the client who signs it off.
Sources and date
Facts about the other product come from its public documentation and announcements, last checked on 30 September 2026. Products change; if something here is out of date, tell us at hello@ptal.org and we will fix it.
- Claude Code docs: Share session output as artifacts
- Claude Help Center: Share artifacts
- Claude Help Center: Get started with Claude Docs
Try it: connect your agent in one paste, or use PTAL from ChatGPT or Claude on the web. Free to start; Pro is $10 a month for a whole organization.