Engage the utility's IT and metering operations teams early; AMI head-end APIs (Itron Enterprise Edition, Landis+Gyr GridStream, etc.) are not publicly documented and require the utility to grant API access and provide vendor-specific documentation under NDA or utility agreement.
Determine the integration protocol the utility's head-end exposes: common options include a REST or SOAP web service, a database replication feed, a file-based export (CSV/XML via SFTP), or a message bus (JMS, Kafka) — each requires a different integration architecture.
For REST/SOAP APIs, obtain the WSDL or OpenAPI specification from the utility; authenticate using the credentials and method the utility provisions (often mutual TLS or IP-restricted API key) and implement the meter read request and interval data retrieval operations.
For file-based exports, negotiate the export schedule, file format (often an ANSI C12.19-derived XML or proprietary CSV), and SFTP drop location; implement a file watcher that parses arrivals, validates record counts against header totals, and loads into your data store.
Test with a representative sample of meter types and reporting scenarios (outages, tamper events, partial intervals) in the utility's staging environment before production; edge cases from firmware differences across meter generations are common and must be handled explicitly.
Known gotchas
AMI head-end APIs are deeply tied to the utility's metering infrastructure version; a utility upgrade to the head-end software can break integrations without notice — establish a change notification process with the utility's metering team as a contractual requirement.
Interval data from AMI systems frequently contains quality flags (e.g., estimated, interpolated, reverse-flow) encoded as numeric codes that vary by vendor and even by utility configuration; ignoring quality flags and treating all intervals as measured data produces incorrect energy calculations.
Head-end systems are typically on isolated operational technology networks with strict firewall rules; network path setup (VPN, proxy, allowlisting) can take weeks to provision through utility security review — start network access requests well before the development timeline requires live data.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp