Today from ~07:40 to ~14:40 UTC, event ingestion into Modem was delayed.
During that window, each customer's context graph stopped updating. Dashboards, the API, and MCP-connected agents reported stale world state, and data-triggered automations ran behind.
No data was lost; the backlog has fully processed and the incident is resolved.
We're sorry. You rely on Modem to be current, and for several hours it wasn't. We'll follow up with a full postmortem and what steps we're taking to mitigate future outages.