Analytics for the surface, not the agent.
Vesta started from a simple problem: people ship MCP servers and agent-facing surfaces, then have no idea how agents actually use them.
A dashboard isn't the answer. A change is.
Surface owners can't see their own users. The agent layer sits in the middle and hides them. You get logs and traces, but not the patterns: which descriptions confuse the model, which tools are dead weight, which call sequences fail.
Vesta reads the behaviour and closes the loop: observe, interpret, recommend, measure. The product is the analysis and the recommendations. Logs and counts are inputs, not the point.
Vesta is built by DataFenix, a data-science consultancy. datafenix.ai.
20 years in data science, 5 of them at Google, fractional Chief AI Officer work today.
Most observability tools are built by infrastructure engineers. The hard part of Vesta is the analysis.
OpenTelemetry-native. One instrument() call, fail-open, runs in your own process.
The product and the people behind it
DataFenix is the data-science consultancy. Vesta is the product built inside it: analytics for MCP servers and agent-facing surfaces. Same people, different shape.
[THE TWO]
One team. Vesta is built by DataFenix.