Connect Claude or ChatGPT to your sprinkler system
Yes — your AI assistant can run your sprinklers. Verdyn's Expert tier is agent-native: your account exposes a personal MCP endpoint, so any MCP-capable assistant — Claude, ChatGPT, or your own agent — can read and manage your watering program the same way the app does. It works the same way you'd connect Claude to any smart-home device, except irrigation is the one place an agent gets real agronomy to reason with: live weather, soil, grass type, and your local watering restrictions.
1 · Get your connector URL
Sign in, open the dashboard, find Connect your AI agent, and press Generate agent key. The URL shown contains your key and is displayed once — treat it like a password. Re-generating or revoking from the same panel cuts off any previously connected agent immediately.
2 · Add it to your assistant
Claude (web or desktop): Settings → Connectors → Add custom connector → paste the URL.
ChatGPT: Settings → Connectors → Add MCP server → paste the URL.
Claude Code / other MCP clients: add it as a streamable-HTTP server URL.
The endpoint is stateless streamable-HTTP (POST JSON-RPC) with the key in the URL path — no OAuth dance required. Scoped OAuth 2.1 authorization is on the roadmap.
3 · Talk to your lawn
“What's my watering plan today, and why?”
“We just laid new sod in the back — set that zone's planting date to today.”
“It's pouring here — pause watering.”
“Did anything actually run this week?”
Available tools
| get_status | controller link, tier, and the full lawn profile |
| get_daily_plan | today's cycles + the plain-English rationale |
| get_week_preview | 7-day schedule preview |
| update_profile | ZIP, soil, address parity, planting date |
| update_zone | per-zone grass, nozzle, sun, slope, area, planting date |
| set_auto_run | pause / resume automatic watering |
| get_run_history | what actually ran, was skipped, or errored |
Every change is validated by the same engine the app uses, and the hard safety gates (legal watering windows, live weather checks) are enforced server-side — an agent can adjust the program, never bypass the guardrails.
Privacy: what your agent can and can't see
Connecting an agent shares and stores no new private data. Your assistant can read exactly what your dashboard shows — this account's lawn profile, schedule, run history, and public weather — and nothing else.
- It cannot see your B-hyve credentials (Verdyn never stores your password), your email, or any other account.
- Verdyn never sees your conversations — the endpoint receives individual tool calls, and we store no prompts or transcripts, only the key's last-used timestamp.
- Nothing is pushed to any AI provider — your assistant pulls from your endpoint only when you ask, and nothing is used for model training.
- Only the key's SHA-256 hash is stored; revoking it in the dashboard cuts off any connected agent instantly.
Full details in the privacy policy.