Add Datadog metrics for MPP Discovery Service MCP

[parvahuja]

Summary

Adds lightweight Datadog metrics for MPP Discovery Service MCP, removes GitHub as the production monitor, and leaves alerting policy to manual Datadog configuration.

What changed

• Adds a small Datadog helper that posts mpp.discovery_mcp.* metrics directly from the Worker.
• Tracks simple runtime HTTP metrics: request count, response duration, route, method, status class, and server errors.
• Adds a one-minute Cloudflare cron health check against https://mpp.dev/mcp/services for the server card, HEAD, initialize, tools/list, get_catalog_status, and search_services.
• Emits catalog count, offer count, cache age, and catalog refresh metrics.
• Deletes the GitHub scheduled health workflow so GitHub Actions is no longer the monitor.
• Removes notification hooks from the manual MCP checker so code emits metrics/check results only.
• Drops the extra request-recorder and MCP-client service modules; the implementation is now flat helper functions.

Validation

• pnpm --filter mpp-services-mcp check
• pnpm check:ci
• pnpm check:types
• pnpm test
• CLOUDFLARE_ACCOUNT_ID=00000000000000000000000000000000 pnpm --filter mpp-services-mcp deploy:dry-run

Rollout notes

Before relying on Datadog dashboards or notifications, configure the production Worker secret DATADOG_API_KEY, deploy the Worker, confirm mpp.discovery_mcp.* metrics arrive in Datadog, then create Datadog notifications manually from those metrics.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
mpp	Ready	Preview, Comment	Jun 24, 2026 1:06am