jevwire
A decision layer for agent ranking, verification and action assessment.
Added to Jevfast
How it uses Jev
jevwire exposes typed Jev judgments through MCP tools, an embeddable decision library, and Claude Code hooks. Its tools cover ranking, candidate selection, claim verification, general evaluation, action gating, and next-step selection.
What Jev decides
Support message triageExample answers · not a recorded Jev response · Source ↗
Question 1 · is urgent
YOUR APP
INSTRUCTION
Does this message convey urgency?
STATE
The sample support message: “Help! My payouts have been failing for 3 days.”
JEV · NOUL
YesNo
Question 2 · department
YOUR APP
INSTRUCTION
Which team should handle this?
STATE
The sample support message: “Help! My payouts have been failing for 3 days.”
JEV · CHOICE
- billing
- technical
- other
Question 3 · frustration
YOUR APP
INSTRUCTION
How frustrated is the sender?
STATE
The sample support message: “Help! My payouts have been failing for 3 days.”
JEV · SCORE
0
App workflow
- Choose an integration
Use the MCP server, call the library functions, or install the Claude Code plugin.
- Supply state and a typed task
Pass the relevant text and bounded decision question to the selected tool or hook.
- Apply the structured result
Use the probabilities and answer fields in the host policy or harness logic.
- Continue or gate the agent
Return a note or denial to the agent, or use the embedded check before allowing a workflow step.
Why it is interesting
The same bounded decision primitives can run as optional tools or be embedded as mandatory harness checks at points before a tool call, after a fetched result, or before a turn ends.