Claude Code artifacts can now call MCP connectors for live, viewer-specific data

Anthropic's Claude Code artifacts can now call MCP connectors, enabling artifacts (dashboards, internal apps, reports) to fetch live data and take actions on demand -- using each viewer's own connected accounts. The Claude Code artifacts documentation now covers the feature in full; the initial announcement came from ClaudeDevs on X.

What's new

Per the artifacts documentation:

• An artifact can call MCP connectors "each time someone views it, so the page shows current data rather than a snapshot from the session that built it."
• Connector calls from artifacts are available on Pro, Max, Team, and Enterprise plans and require Claude Code v2.1.209 or later. On earlier versions, Claude publishes the page with whatever data the session gathered while building it.
• An artifact that calls connectors cannot be shared to a public link on any plan. On Team and Enterprise you can share it within the organization; on Pro and Max, where a public link is the only sharing route, a connector-backed artifact stays private to you.
• Only connectors from your claude.ai account qualify. Local MCP servers configured in Claude Code (such as .mcp.json servers) can supply data while Claude builds the page, but the published page cannot call them.

Why it matters

Artifacts started as a way to publish a live, shareable page from a Claude Code session. Adding MCP connector calls pushes them closer to interactive internal tools:

• Build an artifact once, then let it pull live data when someone opens it. The page fetches on load and can refresh on an interval or via a refresh control.
• Because the artifact uses the viewer's own connectors, creators do not need to re-run sessions to refresh data, and two people opening the same dashboard can see different data depending on what their accounts can access. Per the doc, "The page never sees anyone's credentials; claude.ai makes the calls on the page's behalf."

Permissions, admin controls, and auditing

The documentation answers the governance questions the launch post left open:

• Viewers approve access first. claude.ai asks each viewer for permission before the page's first connector call. A viewer who declines, or who has not connected a connector the page uses, still sees the page without its live sections.
• Actions use the viewer's account too. A page can offer controls that invoke connector tools with side effects, such as posting a message or updating an issue; the action runs through the account of whoever selects the control.
• Pages are scoped at publish time. Claude declares which connectors the page may call as part of publishing, and the page cannot call connectors outside that declaration.
• Org owners hold a kill switch. Connector calls from artifacts have their own admin toggle (Enable artifact connectors under Settings > Capabilities), separate from the Artifacts toggle itself.
• Publishing, sharing, and deleting artifacts are audit-logged under the claude_artifact_* event types on Team and Enterprise plans.

Sources

• Claude Code documentation, "Share session output as artifacts" (the section "Pull live data with MCP connectors"; checked 2026-08-06): https://code.claude.com/docs/en/artifacts
• Launch announcement (July 15, 2026): https://x.com/ClaudeDevs/status/2077489907350856038

Provenance note: as first published (2026-07-27) this page cited only the X announcement. On 2026-08-06 every claim was re-verified against the Claude Code artifacts documentation, which now documents the feature; the version requirement (v2.1.209+), the local-MCP-server limitation, and the permissions/admin/auditing section were added from that doc.