Where Jev sits inside an agent architecture
A taxonomy of repeatable jobs — routing, tool choice, browser action selection, evaluation, guardrails, and agent infrastructure.
1 BUILDS
Browser Agents
Use Jev as the fast decision layer inside observe → act → verify loops.
2 BUILDSAgent Routing
Route tasks, tickets, or subproblems to the right model, agent, or workflow.
3 BUILDSTool Selection
Pick the right tool from a bounded action set without asking a generative model to improvise.
2 BUILDSAgent Evaluation
Judge completed runs, traces, outputs, and patches with typed scores and confidence.
4 BUILDSGuardrails
Gate risky actions and verify outputs before they become irreversible.
3 BUILDSMCP / Skills
Expose Jev judgments as reusable tools that coding agents can call on demand.