> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nelieo.com/llms.txt
> Use this file to discover all available pages before exploring further.

# NSP: Runtime Substrate for AI Agents

> NSP attaches to live applications and exposes their state and actions as structured JSON — giving AI agents >99% accuracy without screenshots.

NSP (Nelieo State Protocol) gives your AI agents direct, structured access to running software. Instead of looking at screenshots and guessing where to click, your agent reads live application state as typed JSON and calls typed functions directly — no UI automation, no pixel coordinates, no guessing.

<CardGroup cols={2}>
  <Card title="Quickstart" icon="rocket" href="/quickstart">
    Install the daemon, configure Chrome, and read your first state snapshot in 5 minutes.
  </Card>

  <Card title="Core Concepts" icon="book" href="/concepts">
    Understand SubstrateState, SSF, actions, reversibility, and how probes work.
  </Card>

  <Card title="API Reference" icon="code" href="/api-reference/overview">
    Complete reference for every REST and WebSocket endpoint on localhost:7842.
  </Card>

  <Card title="Python SDK" icon="python" href="/sdk/python/overview">
    Install nelieo-nsp and start building agents with typed models and async support.
  </Card>
</CardGroup>

***

## Why NSP?

Vision-based agents fail because the interface is broken, not because the AI is weak. At 85% accuracy per action, a 50-step task has a 0.03% end-to-end success rate.

NSP replaces the screenshot loop with direct runtime access:

|                           | NSP        | Vision-Based Agent |
| ------------------------- | ---------- | ------------------ |
| State read latency        | **\~50ms** | 8,000–15,000ms     |
| Action accuracy (JS apps) | **>99%**   | 66–82%             |
| 50-step task completion   | **\~95%**  | \~0.03%            |
| Actions per minute        | **\~120**  | \~4–7              |

***

## How It Works

The NSP daemon (`axon-daemon`) runs on your machine alongside your applications. It has four layers:

<CardGroup cols={2}>
  <Card title="Process Watcher" icon="eye">
    Continuously scans all running OS processes, classifies their runtime (V8 / JVM / CLR / Native), and manages probe lifecycle.
  </Card>

  <Card title="Runtime Probes" icon="cpu">
    Language-specific probes attach to live processes and extract semantic state — without source code, UI automation, or instrumentation.
  </Card>

  <Card title="Semantic Engine" icon="brain">
    Converts raw runtime data into structured SSF JSON with human-readable dot-notation keys.
  </Card>

  <Card title="Agent API" icon="bolt">
    A REST + WebSocket server on localhost:7842 that any AI agent can call to read state and execute actions.
  </Card>
</CardGroup>

***

## Runtime Coverage

<CardGroup cols={2}>
  <Card title="JavaScript / V8" icon="js" href="/probes/v8">
    Gmail, Slack, VS Code, Notion, GitHub Desktop, Salesforce, Discord, and any Electron app. >99% accuracy.
  </Card>

  <Card title="Java / JVM" icon="java" href="/probes/jvm">
    SAP ERP, Eclipse, IntelliJ IDEA, Jenkins, Elasticsearch. >97% accuracy via JVMTI.
  </Card>

  <Card title=".NET / CLR" icon="microsoft" href="/probes/clr">
    Excel, Word, PowerPoint, Visual Studio, WinForms and WPF apps. >97% accuracy via MMF bridge.
  </Card>

  <Card title="Native / C++" icon="c" href="/probes/native">
    Bloomberg Terminal, Photoshop, AutoCAD, Chrome internals. \~75% accuracy via memory grounding.
  </Card>
</CardGroup>

***

## Get Started

<Steps>
  <Step title="Install the Daemon">
    Download and run `Nelieo NSP Setup.exe` from [www.nelieo.com](https://www.nelieo.com/). The installer registers `axon-daemon` as a Windows Service and starts it on port **7842**.
  </Step>

  <Step title="Configure Chrome">
    Relaunch Chrome with `--remote-debugging-port=9222` to enable V8 state extraction for Gmail, Slack, and any web app.
  </Step>

  <Step title="Install the Python SDK">
    ```bash theme={null}
    pip install nelieo-nsp
    ```
  </Step>

  <Step title="Write Your First Agent">
    Attach to any running app by name, read its live state, and execute typed actions — all in a few lines of Python.
  </Step>
</Steps>

<CardGroup cols={2}>
  <Card title="Follow the Quickstart" icon="arrow-right" href="/quickstart">
    Full step-by-step guide with working code examples.
  </Card>

  <Card title="Browse Guides" icon="map" href="/guides/gmail-agent">
    Practical walkthroughs for Gmail, .NET apps, Java apps, and more.
  </Card>
</CardGroup>