learning_ai_common_plat/docs/MCP+A2A/DOMAIN_DASHBOARDS.md

1.6 KiB

Domain — Dashboards (admin-web, tracker-web, ux-lab)

Admin dashboard (dashboards/admin-web)

Existing leverage points

  • It already centralizes many ops capabilities behind UI.
  • It already has a service client layer (src/lib/platform-client.ts) that talks to platform-service.

MCP opportunities

  • Provide “headless equivalents” of the admin UI actions via MCP tools.
  • Use MCP resources to provide the dashboard with richer contextual data:
    • module inventories
    • policy templates
    • incident runbooks

A2A opportunities

  • Build an Ops Copilot that:
    • proposes actions
    • executes via MCP tools
    • links back to the relevant admin dashboard pages

Tracker dashboard (dashboards/tracker-web)

  • Candidate MCP tools (illustrative names; map onto the tracker modules in platform-service such as items/votes/comments/public):
    • tracker.listPublicItems()
    • tracker.submitFeedback()
    • tracker.vote(itemId)

UX Lab (dashboards/ux-lab)

Existing intent

These micro-apps are greenfield UI experiments that intentionally avoid backend dependencies.

MCP opportunity

  • Use MCP as a dataset generator for local-only UX experiments.
    • Example: generate realistic telemetry events & clusters JSON.

Suggested tools

  • dev.generateSampleTelemetryEvents(count, shape, seed) (local-only helper)
  • dev.generateSampleClusters(count, seed) (local-only helper)

Dashboard “component extraction” workflow

An A2A workflow that:

  • identifies patterns in ux-lab (tables/filters/drawers)
  • proposes a migration plan into @bytelyst/dashboard-components
  • auto-generates a PR in the appropriate package