Current limitation

The Port MCP server supports blueprints, entities, actions, and scorecards. It does not support pages or dashboards.

As a result, AI agents such as Claude Code, Cursor, and Windsurf cannot create, update, or delete dashboard pages or widgets programmatically through the MCP interface.

Use case

We built a Claude usage analytics pipeline that syncs data into a Port blueprint. The goal was to create a dashboard with KPI cards, pie charts, and table widgets directly from Claude Code via the MCP server.

We had to fall back to raw REST API calls, which proved brittle. During implementation we encountered undocumented constraints such as:

• Maximum of two dashboard widget sections

• agentIdentifier requirements

• Other implicit validation rules

Requested capabilities

Page management

• create_page to manage dashboard and catalog pages

• update_page to modify existing pages

• delete_page to remove pages

• list_pages to discover existing pages and widget configurations

Widget management

• add_widget to add widgets to a page

• update_widget to modify widgets

• remove_widget to delete widgets

Widget support

• Number charts

• Pie charts

• Table explorers

• Markdown blocks

• Iframe widgets

• Links

Why this matters

As more teams use AI agents to build and maintain internal developer portals, programmatic dashboard composition becomes essential. The REST API exists but is not exposed through MCP. The API also has undocumented limitations that make it harder to use reliably. Exposing page and widget management through MCP would enable stable, fully automated dashboard generation workflows.