Most AI agents rely on fragile vision models or DOM scrapers. They’re slow, expensive, and break the second a UI changes. Nelieo (NSP) is an OS-level Rust daemon that hooks directly into runtime memory (JVM, V8, CLR). It bypasses pixels completely, letting agents read state and execute actions in memory. Key features: • 85ms execution latency • 96% token cost reduction • 100% deterministic (no UI fragility) • Works across Java, V8, and CLRNET apps Pure memory speed for AI agents.

Hey! 👋 I built Nelieo out of frustration. Every time I tried building reliable computer-use agents, they’d fail because an app UI shifted by a few pixels or loaded too slowly. Relying on vision models to click buttons on a screen just felt like the wrong architecture. So I went heads-down into low-level systems engineering to build a Rust daemon that hooks directly into runtime memory (JVM, V8, CLR). Instead of an agent looking at a screen, Nelieo attaches to the process, extracts the raw callable actions in milliseconds, and lets the agent invoke them directly. No visual overhead, 96% fewer tokens, and sub-100ms latency. Would love for you all to check it out, try it on your own stacks, and ask any technical questions!

Nelieo NSP

Strip the UI. Unleash AI agents at full throttle.

Strip the UI. Unleash AI agents at full throttle.