Meet Ray, your new AI coworker.
Ray checks performance, triages alarms, audits contracts and drafts reports, with the evidence behind every answer. It works in Slack, Microsoft Teams and AMMP OS.
On the job before the team logs on
Seven a.m. Ray has already checked yesterday's performance across every site — and knows which issues need a human.
Every month, Ray checks each site against your financial model — and tells bad weather from real problems.
Root-cause analysis, with evidence
When a site underperforms, Ray works out why before anyone opens a spreadsheet.
Weather or site issue?
Ray compares actual output with expected yield. When a site has no irradiance sensor, the platform uses backup satellite irradiance data, so a cloudy month isn't mistaken for a fault.
Neighbouring inverters
Ray compares a device against its neighbouring inverters, for example: INV-03 produced 38% less than its neighbouring inverters.
Across every brand
Because every brand lands in one data model, Ray can investigate underperformance across a mixed fleet, not one portal at a time.
A diagnosis the team can check
Ray hands the team a diagnosis with the evidence behind it. People decide what happens next.
Rank alarms by revenue risk
Not every alarm costs the same. Ray ranks open issues by what they put at risk, so the team fixes the most expensive problem first, not the loudest one.
Priority = (energy at risk × value per kWh + contract penalty) ÷ time to fix
A small fault on a site with a strict availability guarantee can outrank a bigger one that is quick to fix. You see the reasoning behind every ranking, and people keep the decision.
How to rank solar alarms by revenue riskBring your own assistant
Ray runs on the AMMP MCP server. Claude, Copilot or your own agent can use the same portfolio context.
About the MCP server