.png)
OpenAI, read over MCP.
Hatch reads OpenAI agent activity over MCP, so the agent layer is counted the same way the rest of the work is.
How it works.
An invoice tells you what the agents cost. Reading the runs tells you what they returned. Hatch does the second part and puts it beside the first.
Connected once, then nobody thinks about it.
An admin connects the MCP
One admin authorises it. Agent activity then appears in the same picture as everything else Hatch reads.
Hatch starts reading
Hatch reads the work already happening in the tool. Nobody answers a form, nobody fills in a survey, and nothing gets migrated.
CREATE MATERIALIZED VIEW revenue_by_min AS SELECT date_trunc('minute', ts) AS minute, currency, sum(amount) AS revenue, count(*) AS payments FROM payments WHERE status = 'captured' GROUP BY 1, 2;
It joins the picture
The signal sits alongside everything else Hatch reads, including the agent layer, so who is performing and where capacity sits is visible without anyone reporting on it.
curl https://api.basalt.dev/v1/views/revenue_by_min?limit=5 \ -H "Authorization: Bearer $BASALT_TOKEN"
What you get.
What Hatch reads
Agent runs over MCP: what was asked for, and what came back.
Who turns it on
An admin, once. It then runs for everyone without anyone else doing anything.
Who it is for
Founders, CFOs and COOs who are paying for an agent layer and want to see its return.
What it does not ask for
No forms, no surveys, no migration. SOC 2 Type II, and customer data never trains models.
Find out what you actually have.
Thirty days, one named engineer, and a readout of what your team and your agents actually produced. $2,500, credited in full against your first year.