Claude can now officially create whiteboards and spreadsheets.
These two document types—“artifacts” in Anthropic’s jargon—are added, among other things, to the presentations, which were integrated in September.
From one announcement to the next, the message is the same: no need to leave Claude to work on a project… but exporting to third‑party software remains possible.
This positioning, straddling a supporting tool and a substitute, is reinforced by the “interactive connectors.” Under this banner, Anthropic is progressively rolling out the MCP Apps extension. It has reached the point where fully running certain software inside Claude is now possible. For example, Acrobat (Adobe), Confluence (Atlassian) and Notion.
The ‘complement’ logic, also adopted with Claude Data
Whiteboards and spreadsheets are available in beta to paid subscribers. Claude Dashboards, for its part, moves to general availability. But under a different name that reflects its expanded functionality: Claude Data.
Compared to the initial version, which focused on building dashboards, a data storytelling component has been added. Versioning has also been introduced, as well as caching of dashboards and the definition of shared metrics.
It remains, as Anthropic puts it, a “complement” to BI and data analysis tools: dashboards can be transferred to third‑party tools (Amplitude, Mixpanel, Omni, PostHog and Sigma are named).
Compared to SaaS, Anthropic has the edge on inference
The same pattern was followed with Claude Design, a prototyping module launched in April and ultimately merged into the chat in September, at the same time as Claude Cowork. Anthropic isn’t yet at the functional level of specialized software, but it appears capable of attracting a slice of non‑specialists. It’s tempting to see this as a factor behind Figma’s stock decline (about -60% over the past year).
By integrating these features into the Claude subscription, Anthropic has the means to offer users an inference cost lower than what third‑party publishers would require via API. More broadly, it has an opportunity to capture workloads at a time when underlying LLMs are trending toward becoming commodities.
