JEV REGISTRY • VERIFIED TOOL|DEVTOOLS & TERMINAL| 18.9MS EDGE REFLEX
Home/Hub/Terminal Output Evaluator
DevTools & Terminal18.9ms Reflex$0 / MIT License
29,500 active downloads

Terminal Output Evaluator

Deterministic exit code and stdout parser for test and build commands

jevproxy // terminal-command-eval (reflex kernel)
$npx jevproxy run terminal-eval
JevProxy Intercept Resolved(Mode: DIRECT_REFLEX)
18.9ms|Cost: $0.0001|0 reasoning tokens burned
STDOUT • Tool Call Output:Status: 200 OK
agent.dispatch(eval_terminal_output)
payload: { "exit_code": { "type": "number", "description": "Subprocess exit code" }, "stdout": { "ty...
✔ Decision short-circuited in 18.9ms without roundtrip to frontier LLM.
Upstream token bill saved: $0.0210 on this turn.
Terminal Output Evaluator
Compatible with Cursor, Claude Code, Windsurf, OpenCode
TRADITIONAL LLM CALL:UNOPTIMIZED
• Median Latency: 1850 ms
• Cost per Turn: $0.0210
• Mode: Full KV Cache Reload & TTFT Prefill
JEVPROXY REFLEX KERNEL:77x FASTER
• Median Latency: 18.9 ms
• Cost per Turn: $0.0001
• Accuracy: 99.6% deterministic

1-Click CLI Execution

Run this tool accelerated through the JevProxy gateway without manual wiring:

npx jevproxy run terminal-eval

OpenAI / Anthropic Tool Schema

JSON Specification

Paste this schema into your agent tools definition or Cursor extensions:

{
  "type": "function",
  "function": {
    "name": "eval_terminal_output",
    "description": "Parse exit code and stdout/stderr to determine next agent action",
    "parameters": {
      "type": "object",
      "properties": {
        "exit_code": {
          "type": "number",
          "description": "Subprocess exit code"
        },
        "stdout": {
          "type": "string",
          "description": "Standard output snippet"
        },
        "stderr": {
          "type": "string",
          "description": "Standard error snippet"
        }
      },
      "required": [
        "exit_code"
      ]
    }
  }
}

Technical Architecture & Usage

Stop Burning LLM Tokens on Exit Codes

When a test passes with exit code 0, you do not need Claude 3.5 Sonnet to tell you the test passed. This tool classifies the terminal result in 18.9ms and advances the agent immediately.

Accelerate Terminal Output Evaluator with JevProxy

Get 5,000,000 free decision tokens. Eliminate the 3-second tool freeze in Cursor and Claude Code in under 60 seconds.