---
title: "Discussion on \"The Permission Layer Is 98% of Agent Engineering\" | Hashnode"
description: "Discussion on \"The Permission Layer Is 98% of Agent Engineering\". Building an AI coding agent is not primarily about choosing the right model. It's about building the infrastructure around the model that keeps it safe, bounded, and trustworthy. A production agent harness contains only about 1–2% actual AI logic — t..."
url: "https://hashnode.com/posts/the-permission-layer-is-98-of-agent-engineering/69eb7525a2bb5c6c7a1a41ac"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [The Permission Layer Is 98% of Agent Engineering](https://codeongrass.hashnode.dev/the-permission-layer-is-98-of-agent-engineering?utm_source=hashnode&utm_medium=feed)

Building an AI coding agent is not primarily about choosing the right model. It's about building the infrastructure around the model that keeps it safe, bounded, and trustworthy. A production agent harness contains only about 1–2% actual AI logic — t...

Apr 24 · 14 min read · codeongrass.hashnode.dev  
Read the article

-   [#architecture-guide](https://hashnode.com/tag/architecture-guide)
-   [#agent-oversight](https://hashnode.com/tag/agent-oversight)
-   [#claude-code](https://hashnode.com/tag/claude-code)
-   [#grass](https://hashnode.com/tag/grass)
-   [#guide](https://hashnode.com/tag/guide)

## Written by

[Sahil Kathpal](https://hashnode.com/@sahil-kat)

@sahil-kat · Mumbai, India

  
[View profile](https://hashnode.com/@sahil-kat)

Software Engineer building Grass — mobile access for Claude Code and AI coding agents

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
