Jev Use

Jev Use

Overview

Splits an agent loop by whether a step must produce text. Steps that only produce a decision - is the build done, which of these 30 elements to click, is this shell command safe, keep or drop this message - are handed to Jev, TypeSafe's judgment model, which answers typed yes/no, pick-one and score questions about a state in one forward pass instead of generating tokens. The agent stays the planner and the writer; Jev takes the quick calls.

Everything Jev should not decide comes back under a typed escalation contract, so the handoff is explicit in both directions rather than a guess.

When to Use This Skill

• Use when the next step is a judgment over context you already have ("did X succeed?", "which option?", "how bad is this?") rather than something to write.
• Use when you are about to check the same state several times in a row, so the questions can be batched into one call.
• Use when you want a risk check on a proposed tool call before running it.
• Do not use when the step must produce new content - text, code, free-form tool arguments - or when the options cannot be enumerated. Those are structurally the model's own work.

How It Works

Step 1: Install and wire the server

npx -y jev-use install

install wires the stdio MCP server into Claude Code, Codex and pi through each harness's own CLI - whichever it finds - and npx -y jev-use doctor verifies backend resolution with one live round trip. Set a provider credential in the environment the agent runs in (TYPESAFE_API_KEY, OPENROUTER_API_KEY or AI_GATEWAY_API_KEY, auto-detected in that order), or set JEV_BACKEND=mock to run keyless with no network calls.

Step 2: Route each step before working on it

Step 3: Batch every question about one state into one call

jev_judge takes a single state string plus a questions[] array. Latency is flat in question count, and the cost of the shared state amortizes across the batch, so 13 batched questions cost far less than 13 separate calls. Never call it once per question.

Step 4: Honor the escalation contract

Each verdict carries {id, type, answer, confidence, escalate}, plus reason and hint exactly when escalate is true. An escalated verdict is handed back to you - it is a normal verdict with a hint, never an exception:

The two pre-call reasons come from a deterministic router, so a step that was never Jev's does not spend a request.

Examples

Example 1: One state, three questions, one call

// jev_judge input
{
  "state": "CI run #142: build ok, 214 tests passed, 0 failed; 1 test quarantined as flaky last week",
  "questions": [
    { "id": "passed", "type": "noul",   "question": "Did the run fully succeed?" },
    { "id": "next",   "type": "choice", "question": "Next action?",
      "options": { "merge": "everything green", "rerun": "looks flaky", "hold": "needs attention" } },
    { "id": "risk",   "type": "score",  "question": "How risky is merging now?",
      "levels": ["routine", "worth a look", "incident"] }
  ]
}

// result (shape exact, values illustrative)
{
  "verdicts": [
    { "id": "passed", "type": "noul",   "answer": 0.97, "confidence": 0.94, "escalate": false },
    { "id": "next",   "type": "choice", "answer": "merge", "confidence": 0.34, "escalate": true,
      "reason": "unsure",
      "hint": "Treat the answer as a prior, not a decision - reason it out yourself." },
    { "id": "risk",   "type": "score",  "answer": 0.8, "confidence": 0.81, "escalate": false,
      "legend": { "0": "routine", "1": "worth a look", "2": "incident" } }
  ],
  "escalated": true
}

Two verdicts are usable immediately. The third came back escalated with reason: "unsure", so that one question - and only that one - returns to you, with Jev's answer kept as a hint.

A score answer is the expected position on your own levels: 0.8 means "between routine and worth a look, closer to the latter", and legend maps the indices back to your words.

Example 2: Risk check before a tool call

// jev_gate input
{
  "state": "Cleaning up build output in the project checkout after a failed release build",
  "tool": "Bash",
  "input": { "command": "rm -rf ./dist" }
}

// result
{ "decision": "deny", "confidence": 0.88, "hint": "..." }

jev_gate returns allow, deny or escalate for one proposed action. allow is silence: it falls through to the harness's normal permission flow, so the gate can never grant anything - it can only deny or ask. If Jev is unreachable the gate steps aside rather than granting.