[FEATURE] Viewer analytics for Artifacts: views, unique viewers, last-viewed (and expose to the agent)
What I'm trying to do
I publish a recurring report as an Artifact and share the link with a couple of reviewers for sign-off before it goes any further. The Artifact is the review surface: comments on it are the review loop, and that part works well.
The gap
There is no way to tell whether a reviewer has opened the page, and no way for the agent to find out either. The Artifact tool exposes publishing, comments and assets, and nothing about views.
So when reviewers go quiet, I can't distinguish between:
- hasn't opened it yet
- opened it and had no comments
- never got in, because the link didn't work for them
Those need three different follow-ups from me, and right now I have to guess.
Why the third case is more common than it looks
An Artifact is private until it's shared from the share menu, and a viewer without access gets a generic "Page not found" rather than a permission prompt. So a recipient who wasn't granted access reads it as a broken link and often won't come back and ask. Pasting the URL into chat feels like sharing, but isn't, and the failure is silent on both sides.
Viewer analytics would surface that immediately: zero views on a link you believe you shared is a very different signal from views with no comments.
What would be enough
- Per-artifact view count, unique viewers, and last-viewed timestamp
- Ideally, which of the people it's shared with have opened it
- Exposed to the agent as well as the UI, so Claude can answer "has this been looked at yet?" and chase or not chase accordingly
Precedent
Amazon Managed Grafana's Usage Insights gives exactly this per dashboard: views, unique users, and last-active. I use it routinely to tell whether something is actually being read or should be retired. The equivalent question about an Artifact is currently unanswerable, which makes Artifacts harder to trust as a review or hand-off surface than a dashboard.
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗