AIToday
Large Language ModelsAI Business & IndustryZenn AI/MLPublished: Nov 9, 2026, 06:00 JST

EveryInfra: Claude artifact calls skip cache, bill each view at $0.005004

Send on LINE
EveryInfra: Claude artifact calls skip cache, bill each view at $0.005004

3 Key Points

  1. What happened

    EveryInfra's example live dashboard fetched 9 brands' posts in 39 calls (38 charged, 1 refunded) totaling $0.021128, at $0.005004 for 9 calls per view.

  2. Why it matters

    Because the artifact page runtime does not cache the tool's results and cannot auto-refresh them, the page must store results in its own shared storage or viewers trigger repeat paid calls — and the real cost is wait time, with the slowest sync call at 69 seconds.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

Claude added the connector-calling behavior to public artifacts in July 2026, releasing it in Claude Code versions v2.1.207 to v2.1.212. The Help Center's definition is broad: artifacts are things Claude creates and others can view, including documents, slides, designs, dashboards and small tools, and they can be shared, store data and connect to apps.

The economics in EveryInfra's test come from an annotation choice rather than a platform rule. EveryInfra's own data tool, everyinfra_call_api, currently declares readOnlyHint: false and destructiveHint: true. That is the tool's own annotation, not a Claude platform rule for data tools. The practical result is that Claude's page runtime does not cache the tool's results and the page cannot call it through the auto-update watch method, so the page has to carry its own cache.

Cost differences follow from that. At catalog prices checked on 2026-11-05, one view is 9 calls for $0.005004. If 10 people open the page three times a day, that is 900 views and 8,100 billed calls per month without a cache, or $4.50. Saving results to the artifact's shared storage and refreshing every 6 hours brings the same usage to $0.60; a 24-hour shared cache brings it to $0.15.

FAQ
Who pays when someone else opens a published dashboard?
Calls run through the viewer's own connection and access rights, not the publisher's. With an OAuth connector, each person's EveryInfra account is charged; with a shared API key, all costs land on the one account that owns the key.
Why does the page need its own cache?
Because everyinfra_call_api declares readOnlyHint: false and destructiveHint: true, Claude does not cache its results, and the page cannot call it through the auto-update watch method. Without a shared cache, every view triggers paid calls.
Can the dashboard be shared with people outside the organization?
No. Signed-out viewers and people outside the organization do not see the live sections, and 'Anyone with the link' sharing is not available for artifacts that connect to apps.
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleAI firms have sold $300 billion of stock in 2026, with $100 billion more on tap