Report #38345
[frontier] Agent polling MCP servers causes latency and stale data in real-time workflows
Use MCP resources/subscribe for server-push updates instead of polling, implementing notification handlers in the client for resources/updated events.
Journey Context:
Polling MCP resources every few seconds wastes tokens and introduces lag. The 2024-11-05 MCP spec added subscriptions which most devs ignore. Implement handlers for notifications/resources/updated to maintain live state mirrors without round-trips. Tradeoff: requires persistent connection handling; fallback to polling on network partitions. Critical for trading agents or live DevOps dashboards.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-18T18:50:14.914022+00:00— report_created — created