Fleet Status Broadcast

A push-shaped status channel for the fleet — deliberately different from the pull status page (/tools/status.html, which you have to visit). Here the fleet publishes a small machine-readable feed that is regenerated every 5 minutes by an on-host prober (fleet-status-gen.py, systemd timer) that calls every advertised endpoint URL for real — no hand-written claims. Subscribe by polling the feed, reading the RSS edition in any reader, or registering a webhook that fires the moment the overall status changes.

Live right now

loading the feed…

The feeds

JSON   https://api.mekongtradeintel.com/tools/fleet-status.json
RSS    https://api.mekongtradeintel.com/tools/fleet-status.xml
Clinic https://api.mekongtradeintel.com/clinic/status-feed   (same JSON)
Info   https://api.mekongtradeintel.com/clinic/broadcast     (channels + subscriber count)

Poll guidance: the feed regenerates every 5 minutes and is <20 KB — poll at most every 5 minutes. overall is operational or degraded; incidents[] lists anything down with the observed HTTP status. Probe method, stated in the feed itself: each advertised catalogue URL is called with its advertised method (bot UA, so the free-trial layer is skipped and the paid path answers); alive = HTTP 200/400/402/405/422 from a real backend, down = 502/503/504/timeout/connection error. Core surfaces (clinic, MCP connector, tools hub, well-known manifest) are probed the same way.

Webhook push (status changes only)

curl -X POST https://api.mekongtradeintel.com/clinic/broadcast/subscribe \
     -H 'Content-Type: application/json' \
     -d '{"webhook_url": "https://your-agent.example.com/fleet-status"}'

One POST per overall-status change (operational ↔ degraded) — no heartbeats, no per-endpoint spam. Payload: {type:"fleet_status_change", overall, previous_overall, summary, incidents, generated_at, feed}.

Why a broadcast channel?

Buyer agents route around dead endpoints only if they hear about them. A status page helps a human who remembers to look; a feed + RSS + change webhooks let an agent's router react on its own — drop a degraded endpoint, fail over to another slug from endpoint search, or pause a scheduled job until the fleet is operational again.