Perigon turns the live news stream into structured data an application can consume. It monitors around 220,000 sources at roughly 50,000 articles an hour, resolves entities against about 90,000 tracked companies, and serves the result through a machine-readable News API with sub-200ms responses.
The AI-agent angle is genuine rather than decorative. Models are frozen at their training cutoff, and an agent that needs to know what happened this morning needs a live source it can query; Perigon ships an MCP integration so Claude or Cursor can pull from it directly. Monitors handle the other half of the job, watching for defined events and building feeds from them.
This is infrastructure, so the buyer is a developer or a data team rather than an end user. The /pricing page returns a 404, so cost has to come from a conversation, source coverage determines what you can and cannot see, and anything built on a news feed inherits the biases of the outlets in it.
