SLOP State Layer for Observable Programs
A protocol for AI to observe and interact with app state
Start new thread
Diego Carlino

SLOP State Layer for Observable Programs - A protocol for AI to observe and interact with app state

by•
AI agents either parse screenshots (expensive, lossy) or make blind tool calls (no context). SLOP fixes this: apps expose a semantic state tree that AI subscribes to with incremental updates. Actions are contextual: they live on the nodes they affect and appear/disappear as state changes. Ships with a 13-doc spec, 14 SDK packages (TypeScript, Python, Rust, Go), React/Vue/Svelte/Solid/Angular adapters, Chrome extension, desktop app, and CLI. MIT licensed. State-first, not action-first.

Add a comment

Replies

Best
Diego Carlino
Maker
📌
Hey Product Hunt! I built SLOP because I was frustrated with how AI agents interact with apps — either parsing pixels from screenshots or making blind tool calls with zero awareness of what's actually happening. SLOP is state-first: apps publish what they are, AI subscribes and acts in context. The spec is 13 docs, SDKs ship in 4 languages, and there's a Chrome extension that works today. Would love feedback on the protocol design, especially the contextual affordance model vs flat tool registries. Happy to answer any questions!