JEV REGISTRY • VERIFIED TOOL|CODING AGENTS| 16.8MS EDGE REFLEX
Home/Hub/Neurolink
Coding Agents16.8ms Reflex$0 / MIT License
14,820 active downloads

Neurolink

High-speed agent decision streaming bridge for IDE control loops

jevproxy // neurolink (reflex kernel)
$npx jevproxy run neurolink
JevProxy Intercept Resolved(Mode: DIRECT_REFLEX)
16.8ms|Cost: $0.0001|0 reasoning tokens burned
STDOUT • Tool Call Output:Status: 200 OK
agent.dispatch(neurolink_stream_decision)
payload: { "turn_id": { "type": "string", "description": "Current agent turn identifier" }, "action_ty...
✔ Decision short-circuited in 16.8ms without roundtrip to frontier LLM.
Upstream token bill saved: $0.0150 on this turn.
Neurolink
Compatible with Cursor, Claude Code, Windsurf, OpenCode
TRADITIONAL LLM CALL:UNOPTIMIZED
• Median Latency: 1420 ms
• Cost per Turn: $0.0150
• Mode: Full KV Cache Reload & TTFT Prefill
JEVPROXY REFLEX KERNEL:77x FASTER
• Median Latency: 16.8 ms
• Cost per Turn: $0.0001
• Accuracy: 99.4% deterministic

1-Click CLI Execution

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

npx jevproxy run neurolink

OpenAI / Anthropic Tool Schema

JSON Specification

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

{
  "type": "function",
  "function": {
    "name": "neurolink_stream_decision",
    "description": "Stream intermediate agent micro-decision via low-latency IPC socket",
    "parameters": {
      "type": "object",
      "properties": {
        "turn_id": {
          "type": "string",
          "description": "Current agent turn identifier"
        },
        "action_type": {
          "type": "string",
          "enum": [
            "TOOL_DISPATCH",
            "STATE_SYNC",
            "CONTEXT_PRUNE"
          ]
        },
        "payload": {
          "type": "object",
          "description": "Typed action payload"
        }
      },
      "required": [
        "turn_id",
        "action_type"
      ]
    }
  }
}

Technical Architecture & Usage

Overview

Neurolink eliminates the IPC bottleneck in autonomous coding agents. Instead of serializing full conversation histories across HTTP/1.1 connections for trivial tool decisions, Neurolink uses zero-copy binary vector serialization over local sockets.

Architecture

When integrated with JevProxy, Neurolink intercepts client-side tool requests and resolves deterministic branches in 16.8ms. Complex creative prompts bypass the local kernel and stream directly to Claude 3.5 Sonnet or GPT-4o.

Installation & Cursor Setup

# Install Neurolink bridge
npm install -g neurolink-bridge

# Run accelerated by JevProxy
npx jevproxy run neurolink

Accelerate Neurolink with JevProxy

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