Loading
MCP turns your captive-portal data into a tool any LLM agent can call. Built into Claude Desktop, Cursor, Cline, and any MCP-compatible runtime.
Natural-language queries become typed MCP tool calls — no prompt engineering, no schema hunting. The agent discovers tools automatically.
MCP and the REST API share the same auth, rate limits, and data model. The difference is setup time — and who's doing the work.
Copy one of these patterns into Claude Desktop or Cline. The agent figures out the rest — tool discovery is automatic.
Agent runs every Monday, calls get_venue_footfall on every venue in the tenant, computes lowest return-rate locations, and posts a ranked summary to a Slack channel.
// claude.mywifi.io agent config (pseudo-code)
every monday at 08:00
venues = mywifi.list_venues()
for v in venues:
stats = mywifi.get_venue_footfall(venue_id: v.id, period: "7d")
rates.append({ venue: v.name, return_rate: stats.return_rate })
ranked = sort(rates, by: "return_rate", asc: true)
slack.post("#wifi-health", format_digest(ranked))Agent monitors return-rate drops across all venues. When any location drops >5pts week-over-week, it automatically fires an email re-engagement campaign to lapsed guests.
// Threshold watch pattern
every day at 09:00
metrics = mywifi.get_return_rate(period: "7d", compare: "prev_7d")
at_risk = filter(metrics, drop_pts > 5)
for venue in at_risk:
segment = mywifi.query_segment(
venue_id: venue.id,
visits_min: 2,
last_seen_days: 14..90
)
mywifi.fire_campaign(
segment_id: segment.id,
template: "re-engagement-v2",
channel: "whatsapp"
)Agent splits venues 50/50, calls update_portal_theme to deploy variant B to the test group, then monitors conversion rate and return visits for 7 days before reporting.
// A/B test orchestration
venues = mywifi.list_venues()
test_grp = random_sample(venues, pct: 50)
ctrl_grp = venues - test_grp
for v in test_grp:
mywifi.update_portal_theme(
venue_id: v.id,
theme: "variant-b-2026"
)
// After 7d — compare groups
after 7 days:
report = compare_conversion(test_grp, ctrl_grp)
notify_slack("#portal-experiments", report)MCP calls go through the same auth pipeline as the REST API. Tenants control scope. Every action is logged.
Each reseller account has its own scoped API credentials. Agents only see data belonging to the authenticated tenant.
Every tool is labeled READ or WRITE. Tenants can disable all write tools — limiting agents to query-only access.
Every tool call is logged with agent ID, tool name, parameters, and timestamp. Accessible from the reseller dashboard.
MCP tool calls count against the same per-tier rate limit as REST API calls — 1k req/min on MSP. No surprise overages.
Beta is open to MSP-tier resellers. We onboard in batches — fill out the form and we'll reach out within 2 business days.
Available on the MSP tier. Existing MSP customers get beta access on request — just email us.